Spaces:
Sleeping
Sleeping
1st
Browse files- LICENSE +395 -0
- README.md +56 -13
- SECURITY.md +41 -0
- app.py +108 -0
- demo.ipynb +0 -0
- omniparser.py +60 -0
- requirements.txt +16 -0
- util/__init__.py +0 -0
- util/__pycache__/__init__.cpython-312.pyc +0 -0
- util/__pycache__/__init__.cpython-39.pyc +0 -0
- util/__pycache__/action_matching.cpython-39.pyc +0 -0
- util/__pycache__/box_annotator.cpython-312.pyc +0 -0
- util/__pycache__/box_annotator.cpython-39.pyc +0 -0
- util/action_matching.py +425 -0
- util/action_type.py +45 -0
- util/box_annotator.py +262 -0
- utils.py +403 -0
- weights/convert_safetensor_to_pt.py +9 -0
- weights/icon_caption_florence/config.json +237 -0
- weights/icon_caption_florence/generation_config.json +13 -0
- weights/icon_caption_florence/model.safetensors +3 -0
- weights/icon_detect/model.safetensors +3 -0
- weights/icon_detect/model.yaml +127 -0
LICENSE
ADDED
@@ -0,0 +1,395 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Attribution 4.0 International
|
2 |
+
|
3 |
+
=======================================================================
|
4 |
+
|
5 |
+
Creative Commons Corporation ("Creative Commons") is not a law firm and
|
6 |
+
does not provide legal services or legal advice. Distribution of
|
7 |
+
Creative Commons public licenses does not create a lawyer-client or
|
8 |
+
other relationship. Creative Commons makes its licenses and related
|
9 |
+
information available on an "as-is" basis. Creative Commons gives no
|
10 |
+
warranties regarding its licenses, any material licensed under their
|
11 |
+
terms and conditions, or any related information. Creative Commons
|
12 |
+
disclaims all liability for damages resulting from their use to the
|
13 |
+
fullest extent possible.
|
14 |
+
|
15 |
+
Using Creative Commons Public Licenses
|
16 |
+
|
17 |
+
Creative Commons public licenses provide a standard set of terms and
|
18 |
+
conditions that creators and other rights holders may use to share
|
19 |
+
original works of authorship and other material subject to copyright
|
20 |
+
and certain other rights specified in the public license below. The
|
21 |
+
following considerations are for informational purposes only, are not
|
22 |
+
exhaustive, and do not form part of our licenses.
|
23 |
+
|
24 |
+
Considerations for licensors: Our public licenses are
|
25 |
+
intended for use by those authorized to give the public
|
26 |
+
permission to use material in ways otherwise restricted by
|
27 |
+
copyright and certain other rights. Our licenses are
|
28 |
+
irrevocable. Licensors should read and understand the terms
|
29 |
+
and conditions of the license they choose before applying it.
|
30 |
+
Licensors should also secure all rights necessary before
|
31 |
+
applying our licenses so that the public can reuse the
|
32 |
+
material as expected. Licensors should clearly mark any
|
33 |
+
material not subject to the license. This includes other CC-
|
34 |
+
licensed material, or material used under an exception or
|
35 |
+
limitation to copyright. More considerations for licensors:
|
36 |
+
wiki.creativecommons.org/Considerations_for_licensors
|
37 |
+
|
38 |
+
Considerations for the public: By using one of our public
|
39 |
+
licenses, a licensor grants the public permission to use the
|
40 |
+
licensed material under specified terms and conditions. If
|
41 |
+
the licensor's permission is not necessary for any reason--for
|
42 |
+
example, because of any applicable exception or limitation to
|
43 |
+
copyright--then that use is not regulated by the license. Our
|
44 |
+
licenses grant only permissions under copyright and certain
|
45 |
+
other rights that a licensor has authority to grant. Use of
|
46 |
+
the licensed material may still be restricted for other
|
47 |
+
reasons, including because others have copyright or other
|
48 |
+
rights in the material. A licensor may make special requests,
|
49 |
+
such as asking that all changes be marked or described.
|
50 |
+
Although not required by our licenses, you are encouraged to
|
51 |
+
respect those requests where reasonable. More_considerations
|
52 |
+
for the public:
|
53 |
+
wiki.creativecommons.org/Considerations_for_licensees
|
54 |
+
|
55 |
+
=======================================================================
|
56 |
+
|
57 |
+
Creative Commons Attribution 4.0 International Public License
|
58 |
+
|
59 |
+
By exercising the Licensed Rights (defined below), You accept and agree
|
60 |
+
to be bound by the terms and conditions of this Creative Commons
|
61 |
+
Attribution 4.0 International Public License ("Public License"). To the
|
62 |
+
extent this Public License may be interpreted as a contract, You are
|
63 |
+
granted the Licensed Rights in consideration of Your acceptance of
|
64 |
+
these terms and conditions, and the Licensor grants You such rights in
|
65 |
+
consideration of benefits the Licensor receives from making the
|
66 |
+
Licensed Material available under these terms and conditions.
|
67 |
+
|
68 |
+
|
69 |
+
Section 1 -- Definitions.
|
70 |
+
|
71 |
+
a. Adapted Material means material subject to Copyright and Similar
|
72 |
+
Rights that is derived from or based upon the Licensed Material
|
73 |
+
and in which the Licensed Material is translated, altered,
|
74 |
+
arranged, transformed, or otherwise modified in a manner requiring
|
75 |
+
permission under the Copyright and Similar Rights held by the
|
76 |
+
Licensor. For purposes of this Public License, where the Licensed
|
77 |
+
Material is a musical work, performance, or sound recording,
|
78 |
+
Adapted Material is always produced where the Licensed Material is
|
79 |
+
synched in timed relation with a moving image.
|
80 |
+
|
81 |
+
b. Adapter's License means the license You apply to Your Copyright
|
82 |
+
and Similar Rights in Your contributions to Adapted Material in
|
83 |
+
accordance with the terms and conditions of this Public License.
|
84 |
+
|
85 |
+
c. Copyright and Similar Rights means copyright and/or similar rights
|
86 |
+
closely related to copyright including, without limitation,
|
87 |
+
performance, broadcast, sound recording, and Sui Generis Database
|
88 |
+
Rights, without regard to how the rights are labeled or
|
89 |
+
categorized. For purposes of this Public License, the rights
|
90 |
+
specified in Section 2(b)(1)-(2) are not Copyright and Similar
|
91 |
+
Rights.
|
92 |
+
|
93 |
+
d. Effective Technological Measures means those measures that, in the
|
94 |
+
absence of proper authority, may not be circumvented under laws
|
95 |
+
fulfilling obligations under Article 11 of the WIPO Copyright
|
96 |
+
Treaty adopted on December 20, 1996, and/or similar international
|
97 |
+
agreements.
|
98 |
+
|
99 |
+
e. Exceptions and Limitations means fair use, fair dealing, and/or
|
100 |
+
any other exception or limitation to Copyright and Similar Rights
|
101 |
+
that applies to Your use of the Licensed Material.
|
102 |
+
|
103 |
+
f. Licensed Material means the artistic or literary work, database,
|
104 |
+
or other material to which the Licensor applied this Public
|
105 |
+
License.
|
106 |
+
|
107 |
+
g. Licensed Rights means the rights granted to You subject to the
|
108 |
+
terms and conditions of this Public License, which are limited to
|
109 |
+
all Copyright and Similar Rights that apply to Your use of the
|
110 |
+
Licensed Material and that the Licensor has authority to license.
|
111 |
+
|
112 |
+
h. Licensor means the individual(s) or entity(ies) granting rights
|
113 |
+
under this Public License.
|
114 |
+
|
115 |
+
i. Share means to provide material to the public by any means or
|
116 |
+
process that requires permission under the Licensed Rights, such
|
117 |
+
as reproduction, public display, public performance, distribution,
|
118 |
+
dissemination, communication, or importation, and to make material
|
119 |
+
available to the public including in ways that members of the
|
120 |
+
public may access the material from a place and at a time
|
121 |
+
individually chosen by them.
|
122 |
+
|
123 |
+
j. Sui Generis Database Rights means rights other than copyright
|
124 |
+
resulting from Directive 96/9/EC of the European Parliament and of
|
125 |
+
the Council of 11 March 1996 on the legal protection of databases,
|
126 |
+
as amended and/or succeeded, as well as other essentially
|
127 |
+
equivalent rights anywhere in the world.
|
128 |
+
|
129 |
+
k. You means the individual or entity exercising the Licensed Rights
|
130 |
+
under this Public License. Your has a corresponding meaning.
|
131 |
+
|
132 |
+
|
133 |
+
Section 2 -- Scope.
|
134 |
+
|
135 |
+
a. License grant.
|
136 |
+
|
137 |
+
1. Subject to the terms and conditions of this Public License,
|
138 |
+
the Licensor hereby grants You a worldwide, royalty-free,
|
139 |
+
non-sublicensable, non-exclusive, irrevocable license to
|
140 |
+
exercise the Licensed Rights in the Licensed Material to:
|
141 |
+
|
142 |
+
a. reproduce and Share the Licensed Material, in whole or
|
143 |
+
in part; and
|
144 |
+
|
145 |
+
b. produce, reproduce, and Share Adapted Material.
|
146 |
+
|
147 |
+
2. Exceptions and Limitations. For the avoidance of doubt, where
|
148 |
+
Exceptions and Limitations apply to Your use, this Public
|
149 |
+
License does not apply, and You do not need to comply with
|
150 |
+
its terms and conditions.
|
151 |
+
|
152 |
+
3. Term. The term of this Public License is specified in Section
|
153 |
+
6(a).
|
154 |
+
|
155 |
+
4. Media and formats; technical modifications allowed. The
|
156 |
+
Licensor authorizes You to exercise the Licensed Rights in
|
157 |
+
all media and formats whether now known or hereafter created,
|
158 |
+
and to make technical modifications necessary to do so. The
|
159 |
+
Licensor waives and/or agrees not to assert any right or
|
160 |
+
authority to forbid You from making technical modifications
|
161 |
+
necessary to exercise the Licensed Rights, including
|
162 |
+
technical modifications necessary to circumvent Effective
|
163 |
+
Technological Measures. For purposes of this Public License,
|
164 |
+
simply making modifications authorized by this Section 2(a)
|
165 |
+
(4) never produces Adapted Material.
|
166 |
+
|
167 |
+
5. Downstream recipients.
|
168 |
+
|
169 |
+
a. Offer from the Licensor -- Licensed Material. Every
|
170 |
+
recipient of the Licensed Material automatically
|
171 |
+
receives an offer from the Licensor to exercise the
|
172 |
+
Licensed Rights under the terms and conditions of this
|
173 |
+
Public License.
|
174 |
+
|
175 |
+
b. No downstream restrictions. You may not offer or impose
|
176 |
+
any additional or different terms or conditions on, or
|
177 |
+
apply any Effective Technological Measures to, the
|
178 |
+
Licensed Material if doing so restricts exercise of the
|
179 |
+
Licensed Rights by any recipient of the Licensed
|
180 |
+
Material.
|
181 |
+
|
182 |
+
6. No endorsement. Nothing in this Public License constitutes or
|
183 |
+
may be construed as permission to assert or imply that You
|
184 |
+
are, or that Your use of the Licensed Material is, connected
|
185 |
+
with, or sponsored, endorsed, or granted official status by,
|
186 |
+
the Licensor or others designated to receive attribution as
|
187 |
+
provided in Section 3(a)(1)(A)(i).
|
188 |
+
|
189 |
+
b. Other rights.
|
190 |
+
|
191 |
+
1. Moral rights, such as the right of integrity, are not
|
192 |
+
licensed under this Public License, nor are publicity,
|
193 |
+
privacy, and/or other similar personality rights; however, to
|
194 |
+
the extent possible, the Licensor waives and/or agrees not to
|
195 |
+
assert any such rights held by the Licensor to the limited
|
196 |
+
extent necessary to allow You to exercise the Licensed
|
197 |
+
Rights, but not otherwise.
|
198 |
+
|
199 |
+
2. Patent and trademark rights are not licensed under this
|
200 |
+
Public License.
|
201 |
+
|
202 |
+
3. To the extent possible, the Licensor waives any right to
|
203 |
+
collect royalties from You for the exercise of the Licensed
|
204 |
+
Rights, whether directly or through a collecting society
|
205 |
+
under any voluntary or waivable statutory or compulsory
|
206 |
+
licensing scheme. In all other cases the Licensor expressly
|
207 |
+
reserves any right to collect such royalties.
|
208 |
+
|
209 |
+
|
210 |
+
Section 3 -- License Conditions.
|
211 |
+
|
212 |
+
Your exercise of the Licensed Rights is expressly made subject to the
|
213 |
+
following conditions.
|
214 |
+
|
215 |
+
a. Attribution.
|
216 |
+
|
217 |
+
1. If You Share the Licensed Material (including in modified
|
218 |
+
form), You must:
|
219 |
+
|
220 |
+
a. retain the following if it is supplied by the Licensor
|
221 |
+
with the Licensed Material:
|
222 |
+
|
223 |
+
i. identification of the creator(s) of the Licensed
|
224 |
+
Material and any others designated to receive
|
225 |
+
attribution, in any reasonable manner requested by
|
226 |
+
the Licensor (including by pseudonym if
|
227 |
+
designated);
|
228 |
+
|
229 |
+
ii. a copyright notice;
|
230 |
+
|
231 |
+
iii. a notice that refers to this Public License;
|
232 |
+
|
233 |
+
iv. a notice that refers to the disclaimer of
|
234 |
+
warranties;
|
235 |
+
|
236 |
+
v. a URI or hyperlink to the Licensed Material to the
|
237 |
+
extent reasonably practicable;
|
238 |
+
|
239 |
+
b. indicate if You modified the Licensed Material and
|
240 |
+
retain an indication of any previous modifications; and
|
241 |
+
|
242 |
+
c. indicate the Licensed Material is licensed under this
|
243 |
+
Public License, and include the text of, or the URI or
|
244 |
+
hyperlink to, this Public License.
|
245 |
+
|
246 |
+
2. You may satisfy the conditions in Section 3(a)(1) in any
|
247 |
+
reasonable manner based on the medium, means, and context in
|
248 |
+
which You Share the Licensed Material. For example, it may be
|
249 |
+
reasonable to satisfy the conditions by providing a URI or
|
250 |
+
hyperlink to a resource that includes the required
|
251 |
+
information.
|
252 |
+
|
253 |
+
3. If requested by the Licensor, You must remove any of the
|
254 |
+
information required by Section 3(a)(1)(A) to the extent
|
255 |
+
reasonably practicable.
|
256 |
+
|
257 |
+
4. If You Share Adapted Material You produce, the Adapter's
|
258 |
+
License You apply must not prevent recipients of the Adapted
|
259 |
+
Material from complying with this Public License.
|
260 |
+
|
261 |
+
|
262 |
+
Section 4 -- Sui Generis Database Rights.
|
263 |
+
|
264 |
+
Where the Licensed Rights include Sui Generis Database Rights that
|
265 |
+
apply to Your use of the Licensed Material:
|
266 |
+
|
267 |
+
a. for the avoidance of doubt, Section 2(a)(1) grants You the right
|
268 |
+
to extract, reuse, reproduce, and Share all or a substantial
|
269 |
+
portion of the contents of the database;
|
270 |
+
|
271 |
+
b. if You include all or a substantial portion of the database
|
272 |
+
contents in a database in which You have Sui Generis Database
|
273 |
+
Rights, then the database in which You have Sui Generis Database
|
274 |
+
Rights (but not its individual contents) is Adapted Material; and
|
275 |
+
|
276 |
+
c. You must comply with the conditions in Section 3(a) if You Share
|
277 |
+
all or a substantial portion of the contents of the database.
|
278 |
+
|
279 |
+
For the avoidance of doubt, this Section 4 supplements and does not
|
280 |
+
replace Your obligations under this Public License where the Licensed
|
281 |
+
Rights include other Copyright and Similar Rights.
|
282 |
+
|
283 |
+
|
284 |
+
Section 5 -- Disclaimer of Warranties and Limitation of Liability.
|
285 |
+
|
286 |
+
a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE
|
287 |
+
EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS
|
288 |
+
AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF
|
289 |
+
ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,
|
290 |
+
IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,
|
291 |
+
WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
|
292 |
+
PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,
|
293 |
+
ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT
|
294 |
+
KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT
|
295 |
+
ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.
|
296 |
+
|
297 |
+
b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE
|
298 |
+
TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,
|
299 |
+
NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,
|
300 |
+
INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,
|
301 |
+
COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR
|
302 |
+
USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN
|
303 |
+
ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR
|
304 |
+
DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR
|
305 |
+
IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.
|
306 |
+
|
307 |
+
c. The disclaimer of warranties and limitation of liability provided
|
308 |
+
above shall be interpreted in a manner that, to the extent
|
309 |
+
possible, most closely approximates an absolute disclaimer and
|
310 |
+
waiver of all liability.
|
311 |
+
|
312 |
+
|
313 |
+
Section 6 -- Term and Termination.
|
314 |
+
|
315 |
+
a. This Public License applies for the term of the Copyright and
|
316 |
+
Similar Rights licensed here. However, if You fail to comply with
|
317 |
+
this Public License, then Your rights under this Public License
|
318 |
+
terminate automatically.
|
319 |
+
|
320 |
+
b. Where Your right to use the Licensed Material has terminated under
|
321 |
+
Section 6(a), it reinstates:
|
322 |
+
|
323 |
+
1. automatically as of the date the violation is cured, provided
|
324 |
+
it is cured within 30 days of Your discovery of the
|
325 |
+
violation; or
|
326 |
+
|
327 |
+
2. upon express reinstatement by the Licensor.
|
328 |
+
|
329 |
+
For the avoidance of doubt, this Section 6(b) does not affect any
|
330 |
+
right the Licensor may have to seek remedies for Your violations
|
331 |
+
of this Public License.
|
332 |
+
|
333 |
+
c. For the avoidance of doubt, the Licensor may also offer the
|
334 |
+
Licensed Material under separate terms or conditions or stop
|
335 |
+
distributing the Licensed Material at any time; however, doing so
|
336 |
+
will not terminate this Public License.
|
337 |
+
|
338 |
+
d. Sections 1, 5, 6, 7, and 8 survive termination of this Public
|
339 |
+
License.
|
340 |
+
|
341 |
+
|
342 |
+
Section 7 -- Other Terms and Conditions.
|
343 |
+
|
344 |
+
a. The Licensor shall not be bound by any additional or different
|
345 |
+
terms or conditions communicated by You unless expressly agreed.
|
346 |
+
|
347 |
+
b. Any arrangements, understandings, or agreements regarding the
|
348 |
+
Licensed Material not stated herein are separate from and
|
349 |
+
independent of the terms and conditions of this Public License.
|
350 |
+
|
351 |
+
|
352 |
+
Section 8 -- Interpretation.
|
353 |
+
|
354 |
+
a. For the avoidance of doubt, this Public License does not, and
|
355 |
+
shall not be interpreted to, reduce, limit, restrict, or impose
|
356 |
+
conditions on any use of the Licensed Material that could lawfully
|
357 |
+
be made without permission under this Public License.
|
358 |
+
|
359 |
+
b. To the extent possible, if any provision of this Public License is
|
360 |
+
deemed unenforceable, it shall be automatically reformed to the
|
361 |
+
minimum extent necessary to make it enforceable. If the provision
|
362 |
+
cannot be reformed, it shall be severed from this Public License
|
363 |
+
without affecting the enforceability of the remaining terms and
|
364 |
+
conditions.
|
365 |
+
|
366 |
+
c. No term or condition of this Public License will be waived and no
|
367 |
+
failure to comply consented to unless expressly agreed to by the
|
368 |
+
Licensor.
|
369 |
+
|
370 |
+
d. Nothing in this Public License constitutes or may be interpreted
|
371 |
+
as a limitation upon, or waiver of, any privileges and immunities
|
372 |
+
that apply to the Licensor or You, including from the legal
|
373 |
+
processes of any jurisdiction or authority.
|
374 |
+
|
375 |
+
|
376 |
+
=======================================================================
|
377 |
+
|
378 |
+
Creative Commons is not a party to its public
|
379 |
+
licenses. Notwithstanding, Creative Commons may elect to apply one of
|
380 |
+
its public licenses to material it publishes and in those instances
|
381 |
+
will be considered the “Licensor.” The text of the Creative Commons
|
382 |
+
public licenses is dedicated to the public domain under the CC0 Public
|
383 |
+
Domain Dedication. Except for the limited purpose of indicating that
|
384 |
+
material is shared under a Creative Commons public license or as
|
385 |
+
otherwise permitted by the Creative Commons policies published at
|
386 |
+
creativecommons.org/policies, Creative Commons does not authorize the
|
387 |
+
use of the trademark "Creative Commons" or any other trademark or logo
|
388 |
+
of Creative Commons without its prior written consent including,
|
389 |
+
without limitation, in connection with any unauthorized modifications
|
390 |
+
to any of its public licenses or any other arrangements,
|
391 |
+
understandings, or agreements concerning use of licensed material. For
|
392 |
+
the avoidance of doubt, this paragraph does not form part of the
|
393 |
+
public licenses.
|
394 |
+
|
395 |
+
Creative Commons may be contacted at creativecommons.org.
|
README.md
CHANGED
@@ -1,13 +1,56 @@
|
|
1 |
-
|
2 |
-
|
3 |
-
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# OmniParser: Screen Parsing tool for Pure Vision Based GUI Agent
|
2 |
+
|
3 |
+
<p align="center">
|
4 |
+
<img src="imgs/logo.png" alt="Logo">
|
5 |
+
</p>
|
6 |
+
|
7 |
+
[![arXiv](https://img.shields.io/badge/Paper-green)](https://arxiv.org/abs/2408.00203)
|
8 |
+
[![License](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
|
9 |
+
|
10 |
+
📢 [[Project Page](https://microsoft.github.io/OmniParser/)] [[Blog Post](https://www.microsoft.com/en-us/research/articles/omniparser-for-pure-vision-based-gui-agent/)] [[Models](https://huggingface.co/microsoft/OmniParser)]
|
11 |
+
|
12 |
+
**OmniParser** is a comprehensive method for parsing user interface screenshots into structured and easy-to-understand elements, which significantly enhances the ability of GPT-4V to generate actions that can be accurately grounded in the corresponding regions of the interface.
|
13 |
+
|
14 |
+
## News
|
15 |
+
- [2024/10] Both Interactive Region Detection Model and Icon functional description model are released! [Hugginface models](https://huggingface.co/microsoft/OmniParser)
|
16 |
+
- [2024/09] OmniParser achieves the best performance on [Windows Agent Arena](https://microsoft.github.io/WindowsAgentArena/)!
|
17 |
+
|
18 |
+
## Install
|
19 |
+
Install environment:
|
20 |
+
```python
|
21 |
+
conda create -n "omni" python==3.12
|
22 |
+
conda activate omni
|
23 |
+
pip install -r requirements.txt
|
24 |
+
```
|
25 |
+
|
26 |
+
Then download the model ckpts files in: https://huggingface.co/microsoft/OmniParser, and put them under weights/, default folder structure is: weights/icon_detect, weights/icon_caption_florence, weights/icon_caption_blip2.
|
27 |
+
|
28 |
+
Finally, convert the safetensor to .pt file.
|
29 |
+
```python
|
30 |
+
python weights/convert_safetensor_to_pt.py
|
31 |
+
```
|
32 |
+
|
33 |
+
## Examples:
|
34 |
+
We put together a few simple examples in the demo.ipynb.
|
35 |
+
|
36 |
+
## Gradio Demo
|
37 |
+
To run gradio demo, simply run:
|
38 |
+
```python
|
39 |
+
python gradio_demo.py
|
40 |
+
```
|
41 |
+
|
42 |
+
|
43 |
+
## 📚 Citation
|
44 |
+
Our technical report can be found [here](https://arxiv.org/abs/2408.00203).
|
45 |
+
If you find our work useful, please consider citing our work:
|
46 |
+
```
|
47 |
+
@misc{lu2024omniparserpurevisionbased,
|
48 |
+
title={OmniParser for Pure Vision Based GUI Agent},
|
49 |
+
author={Yadong Lu and Jianwei Yang and Yelong Shen and Ahmed Awadallah},
|
50 |
+
year={2024},
|
51 |
+
eprint={2408.00203},
|
52 |
+
archivePrefix={arXiv},
|
53 |
+
primaryClass={cs.CV},
|
54 |
+
url={https://arxiv.org/abs/2408.00203},
|
55 |
+
}
|
56 |
+
```
|
SECURITY.md
ADDED
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<!-- BEGIN MICROSOFT SECURITY.MD V0.0.9 BLOCK -->
|
2 |
+
|
3 |
+
## Security
|
4 |
+
|
5 |
+
Microsoft takes the security of our software products and services seriously, which includes all source code repositories managed through our GitHub organizations, which include [Microsoft](https://github.com/Microsoft), [Azure](https://github.com/Azure), [DotNet](https://github.com/dotnet), [AspNet](https://github.com/aspnet) and [Xamarin](https://github.com/xamarin).
|
6 |
+
|
7 |
+
If you believe you have found a security vulnerability in any Microsoft-owned repository that meets [Microsoft's definition of a security vulnerability](https://aka.ms/security.md/definition), please report it to us as described below.
|
8 |
+
|
9 |
+
## Reporting Security Issues
|
10 |
+
|
11 |
+
**Please do not report security vulnerabilities through public GitHub issues.**
|
12 |
+
|
13 |
+
Instead, please report them to the Microsoft Security Response Center (MSRC) at [https://msrc.microsoft.com/create-report](https://aka.ms/security.md/msrc/create-report).
|
14 |
+
|
15 |
+
If you prefer to submit without logging in, send email to [secure@microsoft.com](mailto:secure@microsoft.com). If possible, encrypt your message with our PGP key; please download it from the [Microsoft Security Response Center PGP Key page](https://aka.ms/security.md/msrc/pgp).
|
16 |
+
|
17 |
+
You should receive a response within 24 hours. If for some reason you do not, please follow up via email to ensure we received your original message. Additional information can be found at [microsoft.com/msrc](https://www.microsoft.com/msrc).
|
18 |
+
|
19 |
+
Please include the requested information listed below (as much as you can provide) to help us better understand the nature and scope of the possible issue:
|
20 |
+
|
21 |
+
* Type of issue (e.g. buffer overflow, SQL injection, cross-site scripting, etc.)
|
22 |
+
* Full paths of source file(s) related to the manifestation of the issue
|
23 |
+
* The location of the affected source code (tag/branch/commit or direct URL)
|
24 |
+
* Any special configuration required to reproduce the issue
|
25 |
+
* Step-by-step instructions to reproduce the issue
|
26 |
+
* Proof-of-concept or exploit code (if possible)
|
27 |
+
* Impact of the issue, including how an attacker might exploit the issue
|
28 |
+
|
29 |
+
This information will help us triage your report more quickly.
|
30 |
+
|
31 |
+
If you are reporting for a bug bounty, more complete reports can contribute to a higher bounty award. Please visit our [Microsoft Bug Bounty Program](https://aka.ms/security.md/msrc/bounty) page for more details about our active programs.
|
32 |
+
|
33 |
+
## Preferred Languages
|
34 |
+
|
35 |
+
We prefer all communications to be in English.
|
36 |
+
|
37 |
+
## Policy
|
38 |
+
|
39 |
+
Microsoft follows the principle of [Coordinated Vulnerability Disclosure](https://aka.ms/security.md/cvd).
|
40 |
+
|
41 |
+
<!-- END MICROSOFT SECURITY.MD BLOCK -->
|
app.py
ADDED
@@ -0,0 +1,108 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Optional
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
import numpy as np
|
5 |
+
import torch
|
6 |
+
from PIL import Image
|
7 |
+
import io
|
8 |
+
|
9 |
+
|
10 |
+
import base64, os
|
11 |
+
from utils import check_ocr_box, get_yolo_model, get_caption_model_processor, get_som_labeled_img
|
12 |
+
import torch
|
13 |
+
from PIL import Image
|
14 |
+
|
15 |
+
yolo_model = get_yolo_model(model_path='weights/icon_detect/best.pt')
|
16 |
+
caption_model_processor = get_caption_model_processor(model_name="florence2", model_name_or_path="weights/icon_caption_florence")
|
17 |
+
platform = 'pc'
|
18 |
+
if platform == 'pc':
|
19 |
+
draw_bbox_config = {
|
20 |
+
'text_scale': 0.8,
|
21 |
+
'text_thickness': 2,
|
22 |
+
'text_padding': 2,
|
23 |
+
'thickness': 2,
|
24 |
+
}
|
25 |
+
elif platform == 'web':
|
26 |
+
draw_bbox_config = {
|
27 |
+
'text_scale': 0.8,
|
28 |
+
'text_thickness': 2,
|
29 |
+
'text_padding': 3,
|
30 |
+
'thickness': 3,
|
31 |
+
}
|
32 |
+
elif platform == 'mobile':
|
33 |
+
draw_bbox_config = {
|
34 |
+
'text_scale': 0.8,
|
35 |
+
'text_thickness': 2,
|
36 |
+
'text_padding': 3,
|
37 |
+
'thickness': 3,
|
38 |
+
}
|
39 |
+
|
40 |
+
|
41 |
+
|
42 |
+
MARKDOWN = """
|
43 |
+
# OmniParser for Pure Vision Based General GUI Agent 🔥
|
44 |
+
<div>
|
45 |
+
<a href="https://arxiv.org/pdf/2408.00203">
|
46 |
+
<img src="https://img.shields.io/badge/arXiv-2408.00203-b31b1b.svg" alt="Arxiv" style="display:inline-block;">
|
47 |
+
</a>
|
48 |
+
</div>
|
49 |
+
|
50 |
+
OmniParser is a screen parsing tool to convert general GUI screen to structured elements.
|
51 |
+
"""
|
52 |
+
|
53 |
+
DEVICE = torch.device('cuda')
|
54 |
+
|
55 |
+
# @spaces.GPU
|
56 |
+
# @torch.inference_mode()
|
57 |
+
# @torch.autocast(device_type="cuda", dtype=torch.bfloat16)
|
58 |
+
def process(
|
59 |
+
image_input,
|
60 |
+
box_threshold,
|
61 |
+
iou_threshold
|
62 |
+
) -> Optional[Image.Image]:
|
63 |
+
|
64 |
+
image_save_path = 'imgs/saved_image_demo.png'
|
65 |
+
image_input.save(image_save_path)
|
66 |
+
# import pdb; pdb.set_trace()
|
67 |
+
|
68 |
+
ocr_bbox_rslt, is_goal_filtered = check_ocr_box(image_save_path, display_img = False, output_bb_format='xyxy', goal_filtering=None, easyocr_args={'paragraph': False, 'text_threshold':0.9})
|
69 |
+
text, ocr_bbox = ocr_bbox_rslt
|
70 |
+
# print('prompt:', prompt)
|
71 |
+
dino_labled_img, label_coordinates, parsed_content_list = get_som_labeled_img(image_save_path, yolo_model, BOX_TRESHOLD = box_threshold, output_coord_in_ratio=True, ocr_bbox=ocr_bbox,draw_bbox_config=draw_bbox_config, caption_model_processor=caption_model_processor, ocr_text=text,iou_threshold=iou_threshold)
|
72 |
+
image = Image.open(io.BytesIO(base64.b64decode(dino_labled_img)))
|
73 |
+
print('finish processing')
|
74 |
+
parsed_content_list = '\n'.join(parsed_content_list)
|
75 |
+
return image, str(parsed_content_list)
|
76 |
+
|
77 |
+
|
78 |
+
|
79 |
+
with gr.Blocks() as demo:
|
80 |
+
gr.Markdown(MARKDOWN)
|
81 |
+
with gr.Row():
|
82 |
+
with gr.Column():
|
83 |
+
image_input_component = gr.Image(
|
84 |
+
type='pil', label='Upload image')
|
85 |
+
# set the threshold for removing the bounding boxes with low confidence, default is 0.05
|
86 |
+
box_threshold_component = gr.Slider(
|
87 |
+
label='Box Threshold', minimum=0.01, maximum=1.0, step=0.01, value=0.05)
|
88 |
+
# set the threshold for removing the bounding boxes with large overlap, default is 0.1
|
89 |
+
iou_threshold_component = gr.Slider(
|
90 |
+
label='IOU Threshold', minimum=0.01, maximum=1.0, step=0.01, value=0.1)
|
91 |
+
submit_button_component = gr.Button(
|
92 |
+
value='Submit', variant='primary')
|
93 |
+
with gr.Column():
|
94 |
+
image_output_component = gr.Image(type='pil', label='Image Output')
|
95 |
+
text_output_component = gr.Textbox(label='Parsed screen elements', placeholder='Text Output')
|
96 |
+
|
97 |
+
submit_button_component.click(
|
98 |
+
fn=process,
|
99 |
+
inputs=[
|
100 |
+
image_input_component,
|
101 |
+
box_threshold_component,
|
102 |
+
iou_threshold_component
|
103 |
+
],
|
104 |
+
outputs=[image_output_component, text_output_component]
|
105 |
+
)
|
106 |
+
|
107 |
+
# demo.launch(debug=False, show_error=True, share=True)
|
108 |
+
demo.launch(share=True, server_port=7861, server_name='0.0.0.0')
|
demo.ipynb
ADDED
The diff for this file is too large to render.
See raw diff
|
|
omniparser.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from utils import get_som_labeled_img, check_ocr_box, get_caption_model_processor, get_dino_model, get_yolo_model
|
2 |
+
import torch
|
3 |
+
from ultralytics import YOLO
|
4 |
+
from PIL import Image
|
5 |
+
from typing import Dict, Tuple, List
|
6 |
+
import io
|
7 |
+
import base64
|
8 |
+
|
9 |
+
|
10 |
+
config = {
|
11 |
+
'som_model_path': 'finetuned_icon_detect.pt',
|
12 |
+
'device': 'cpu',
|
13 |
+
'caption_model_path': 'Salesforce/blip2-opt-2.7b',
|
14 |
+
'draw_bbox_config': {
|
15 |
+
'text_scale': 0.8,
|
16 |
+
'text_thickness': 2,
|
17 |
+
'text_padding': 3,
|
18 |
+
'thickness': 3,
|
19 |
+
},
|
20 |
+
'BOX_TRESHOLD': 0.05
|
21 |
+
}
|
22 |
+
|
23 |
+
|
24 |
+
class Omniparser(object):
|
25 |
+
def __init__(self, config: Dict):
|
26 |
+
self.config = config
|
27 |
+
|
28 |
+
self.som_model = get_yolo_model(model_path=config['som_model_path'])
|
29 |
+
# self.caption_model_processor = get_caption_model_processor(config['caption_model_path'], device=cofig['device'])
|
30 |
+
# self.caption_model_processor['model'].to(torch.float32)
|
31 |
+
|
32 |
+
def parse(self, image_path: str):
|
33 |
+
print('Parsing image:', image_path)
|
34 |
+
ocr_bbox_rslt, is_goal_filtered = check_ocr_box(image_path, display_img = False, output_bb_format='xyxy', goal_filtering=None, easyocr_args={'paragraph': False, 'text_threshold':0.9})
|
35 |
+
text, ocr_bbox = ocr_bbox_rslt
|
36 |
+
|
37 |
+
draw_bbox_config = self.config['draw_bbox_config']
|
38 |
+
BOX_TRESHOLD = self.config['BOX_TRESHOLD']
|
39 |
+
dino_labled_img, label_coordinates, parsed_content_list = get_som_labeled_img(image_path, self.som_model, BOX_TRESHOLD = BOX_TRESHOLD, output_coord_in_ratio=False, ocr_bbox=ocr_bbox,draw_bbox_config=draw_bbox_config, caption_model_processor=None, ocr_text=text,use_local_semantics=False)
|
40 |
+
|
41 |
+
image = Image.open(io.BytesIO(base64.b64decode(dino_labled_img)))
|
42 |
+
# formating output
|
43 |
+
return_list = [{'from': 'omniparser', 'shape': {'x':coord[0], 'y':coord[1], 'width':coord[2], 'height':coord[3]},
|
44 |
+
'text': parsed_content_list[i].split(': ')[1], 'type':'text'} for i, (k, coord) in enumerate(label_coordinates.items()) if i < len(parsed_content_list)]
|
45 |
+
return_list.extend(
|
46 |
+
[{'from': 'omniparser', 'shape': {'x':coord[0], 'y':coord[1], 'width':coord[2], 'height':coord[3]},
|
47 |
+
'text': 'None', 'type':'icon'} for i, (k, coord) in enumerate(label_coordinates.items()) if i >= len(parsed_content_list)]
|
48 |
+
)
|
49 |
+
|
50 |
+
return [image, return_list]
|
51 |
+
|
52 |
+
parser = Omniparser(config)
|
53 |
+
image_path = 'examples/pc_1.png'
|
54 |
+
|
55 |
+
# time the parser
|
56 |
+
import time
|
57 |
+
s = time.time()
|
58 |
+
image, parsed_content_list = parser.parse(image_path)
|
59 |
+
device = config['device']
|
60 |
+
print(f'Time taken for Omniparser on {device}:', time.time() - s)
|
requirements.txt
ADDED
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
easyocr
|
3 |
+
torchvision
|
4 |
+
supervision==0.18.0
|
5 |
+
openai==1.3.5
|
6 |
+
transformers
|
7 |
+
ultralytics==8.1.24
|
8 |
+
azure-identity
|
9 |
+
numpy
|
10 |
+
opencv-python
|
11 |
+
opencv-python-headless
|
12 |
+
gradio
|
13 |
+
dill
|
14 |
+
accelerate
|
15 |
+
timm
|
16 |
+
einops==0.8.0
|
util/__init__.py
ADDED
File without changes
|
util/__pycache__/__init__.cpython-312.pyc
ADDED
Binary file (139 Bytes). View file
|
|
util/__pycache__/__init__.cpython-39.pyc
ADDED
Binary file (141 Bytes). View file
|
|
util/__pycache__/action_matching.cpython-39.pyc
ADDED
Binary file (8.49 kB). View file
|
|
util/__pycache__/box_annotator.cpython-312.pyc
ADDED
Binary file (9.79 kB). View file
|
|
util/__pycache__/box_annotator.cpython-39.pyc
ADDED
Binary file (6.57 kB). View file
|
|
util/action_matching.py
ADDED
@@ -0,0 +1,425 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
'''
|
2 |
+
Adapted from https://github.com/google-research/google-research/tree/master/android_in_the_wild
|
3 |
+
'''
|
4 |
+
|
5 |
+
import jax
|
6 |
+
import jax.numpy as jnp
|
7 |
+
import numpy as np
|
8 |
+
|
9 |
+
# import action_type as action_type_lib
|
10 |
+
import enum
|
11 |
+
|
12 |
+
class ActionType(enum.IntEnum):
|
13 |
+
# Placeholders for unused enum values
|
14 |
+
UNUSED_0 = 0
|
15 |
+
UNUSED_1 = 1
|
16 |
+
UNUSED_2 = 2
|
17 |
+
UNUSED_8 = 8
|
18 |
+
UNUSED_9 = 9
|
19 |
+
|
20 |
+
########### Agent actions ###########
|
21 |
+
|
22 |
+
# A type action that sends text to the emulator. Note that this simply sends
|
23 |
+
# text and does not perform any clicks for element focus or enter presses for
|
24 |
+
# submitting text.
|
25 |
+
TYPE = 3
|
26 |
+
|
27 |
+
# The dual point action used to represent all gestures.
|
28 |
+
DUAL_POINT = 4
|
29 |
+
|
30 |
+
# These actions differentiate pressing the home and back button from touches.
|
31 |
+
# They represent explicit presses of back and home performed using ADB.
|
32 |
+
PRESS_BACK = 5
|
33 |
+
PRESS_HOME = 6
|
34 |
+
|
35 |
+
# An action representing that ADB command for hitting enter was performed.
|
36 |
+
PRESS_ENTER = 7
|
37 |
+
|
38 |
+
########### Episode status actions ###########
|
39 |
+
|
40 |
+
# An action used to indicate the desired task has been completed and resets
|
41 |
+
# the environment. This action should also be used in the case that the task
|
42 |
+
# has already been completed and there is nothing to do.
|
43 |
+
# e.g. The task is to turn on the Wi-Fi when it is already on
|
44 |
+
STATUS_TASK_COMPLETE = 10
|
45 |
+
|
46 |
+
# An action used to indicate that desired task is impossible to complete and
|
47 |
+
# resets the environment. This can be a result of many different things
|
48 |
+
# including UI changes, Android version differences, etc.
|
49 |
+
STATUS_TASK_IMPOSSIBLE = 11
|
50 |
+
|
51 |
+
|
52 |
+
_TAP_DISTANCE_THRESHOLD = 0.14 # Fraction of the screen
|
53 |
+
ANNOTATION_WIDTH_AUGMENT_FRACTION = 1.4
|
54 |
+
ANNOTATION_HEIGHT_AUGMENT_FRACTION = 1.4
|
55 |
+
|
56 |
+
# Interval determining if an action is a tap or a swipe.
|
57 |
+
_SWIPE_DISTANCE_THRESHOLD = 0.04
|
58 |
+
|
59 |
+
|
60 |
+
def _yx_in_bounding_boxes(
|
61 |
+
yx, bounding_boxes
|
62 |
+
):
|
63 |
+
"""Check if the (y,x) point is contained in each bounding box.
|
64 |
+
|
65 |
+
Args:
|
66 |
+
yx: The (y, x) coordinate in pixels of the point.
|
67 |
+
bounding_boxes: A 2D int array of shape (num_bboxes, 4), where each row
|
68 |
+
represents a bounding box: (y_top_left, x_top_left, box_height,
|
69 |
+
box_width). Note: containment is inclusive of the bounding box edges.
|
70 |
+
|
71 |
+
Returns:
|
72 |
+
is_inside: A 1D bool array where each element specifies if the point is
|
73 |
+
contained within the respective box.
|
74 |
+
"""
|
75 |
+
y, x = yx
|
76 |
+
|
77 |
+
# `bounding_boxes` has shape (n_elements, 4); we extract each array along the
|
78 |
+
# last axis into shape (n_elements, 1), then squeeze unneeded dimension.
|
79 |
+
top, left, height, width = [
|
80 |
+
jnp.squeeze(v, axis=-1) for v in jnp.split(bounding_boxes, 4, axis=-1)
|
81 |
+
]
|
82 |
+
|
83 |
+
# The y-axis is inverted for AndroidEnv, so bottom = top + height.
|
84 |
+
bottom, right = top + height, left + width
|
85 |
+
|
86 |
+
return jnp.logical_and(y >= top, y <= bottom) & jnp.logical_and(
|
87 |
+
x >= left, x <= right)
|
88 |
+
|
89 |
+
|
90 |
+
def _resize_annotation_bounding_boxes(
|
91 |
+
annotation_positions, annotation_width_augment_fraction,
|
92 |
+
annotation_height_augment_fraction):
|
93 |
+
"""Resize the bounding boxes by the given fractions.
|
94 |
+
|
95 |
+
Args:
|
96 |
+
annotation_positions: Array of shape (N, 4), where each row represents the
|
97 |
+
(y, x, height, width) of the bounding boxes.
|
98 |
+
annotation_width_augment_fraction: The fraction to augment the box widths,
|
99 |
+
E.g., 1.4 == 240% total increase.
|
100 |
+
annotation_height_augment_fraction: Same as described for width, but for box
|
101 |
+
height.
|
102 |
+
|
103 |
+
Returns:
|
104 |
+
Resized bounding box.
|
105 |
+
|
106 |
+
"""
|
107 |
+
height_change = (
|
108 |
+
annotation_height_augment_fraction * annotation_positions[:, 2])
|
109 |
+
width_change = (
|
110 |
+
annotation_width_augment_fraction * annotation_positions[:, 3])
|
111 |
+
|
112 |
+
# Limit bounding box positions to the screen.
|
113 |
+
resized_annotations = jnp.stack([
|
114 |
+
jnp.maximum(0, annotation_positions[:, 0] - (height_change / 2)),
|
115 |
+
jnp.maximum(0, annotation_positions[:, 1] - (width_change / 2)),
|
116 |
+
jnp.minimum(1, annotation_positions[:, 2] + height_change),
|
117 |
+
jnp.minimum(1, annotation_positions[:, 3] + width_change),
|
118 |
+
],
|
119 |
+
axis=1)
|
120 |
+
return resized_annotations
|
121 |
+
|
122 |
+
|
123 |
+
def is_tap_action(normalized_start_yx,
|
124 |
+
normalized_end_yx):
|
125 |
+
distance = jnp.linalg.norm(
|
126 |
+
jnp.array(normalized_start_yx) - jnp.array(normalized_end_yx))
|
127 |
+
return distance <= _SWIPE_DISTANCE_THRESHOLD
|
128 |
+
|
129 |
+
|
130 |
+
def _is_non_dual_point_action(action_type):
|
131 |
+
return jnp.not_equal(action_type, ActionType.DUAL_POINT)
|
132 |
+
|
133 |
+
|
134 |
+
def _check_tap_actions_match(
|
135 |
+
tap_1_yx,
|
136 |
+
tap_2_yx,
|
137 |
+
annotation_positions,
|
138 |
+
matching_tap_distance_threshold_screen_percentage,
|
139 |
+
annotation_width_augment_fraction,
|
140 |
+
annotation_height_augment_fraction,
|
141 |
+
):
|
142 |
+
"""Determines if two tap actions are the same."""
|
143 |
+
resized_annotation_positions = _resize_annotation_bounding_boxes(
|
144 |
+
annotation_positions,
|
145 |
+
annotation_width_augment_fraction,
|
146 |
+
annotation_height_augment_fraction,
|
147 |
+
)
|
148 |
+
|
149 |
+
# Check if the ground truth tap action falls in an annotation's bounding box.
|
150 |
+
tap1_in_box = _yx_in_bounding_boxes(tap_1_yx, resized_annotation_positions)
|
151 |
+
tap2_in_box = _yx_in_bounding_boxes(tap_2_yx, resized_annotation_positions)
|
152 |
+
both_in_box = jnp.max(tap1_in_box & tap2_in_box)
|
153 |
+
|
154 |
+
# If the ground-truth tap action falls outside any of the annotation
|
155 |
+
# bounding boxes or one of the actions is inside a bounding box and the other
|
156 |
+
# is outside bounding box or vice versa, compare the points using Euclidean
|
157 |
+
# distance.
|
158 |
+
within_threshold = (
|
159 |
+
jnp.linalg.norm(jnp.array(tap_1_yx) - jnp.array(tap_2_yx))
|
160 |
+
<= matching_tap_distance_threshold_screen_percentage
|
161 |
+
)
|
162 |
+
return jnp.logical_or(both_in_box, within_threshold)
|
163 |
+
|
164 |
+
|
165 |
+
def _check_drag_actions_match(
|
166 |
+
drag_1_touch_yx,
|
167 |
+
drag_1_lift_yx,
|
168 |
+
drag_2_touch_yx,
|
169 |
+
drag_2_lift_yx,
|
170 |
+
):
|
171 |
+
"""Determines if two drag actions are the same."""
|
172 |
+
# Store drag deltas (the change in the y and x coordinates from touch to
|
173 |
+
# lift), magnitudes, and the index of the main axis, which is the axis with
|
174 |
+
# the greatest change in coordinate value (e.g. a drag starting at (0, 0) and
|
175 |
+
# ending at (0.3, 0.5) has a main axis index of 1).
|
176 |
+
drag_1_deltas = drag_1_lift_yx - drag_1_touch_yx
|
177 |
+
drag_1_magnitudes = jnp.abs(drag_1_deltas)
|
178 |
+
drag_1_main_axis = np.argmax(drag_1_magnitudes)
|
179 |
+
drag_2_deltas = drag_2_lift_yx - drag_2_touch_yx
|
180 |
+
drag_2_magnitudes = jnp.abs(drag_2_deltas)
|
181 |
+
drag_2_main_axis = np.argmax(drag_2_magnitudes)
|
182 |
+
|
183 |
+
return jnp.equal(drag_1_main_axis, drag_2_main_axis)
|
184 |
+
|
185 |
+
|
186 |
+
def check_actions_match(
|
187 |
+
action_1_touch_yx,
|
188 |
+
action_1_lift_yx,
|
189 |
+
action_1_action_type,
|
190 |
+
action_2_touch_yx,
|
191 |
+
action_2_lift_yx,
|
192 |
+
action_2_action_type,
|
193 |
+
annotation_positions,
|
194 |
+
tap_distance_threshold = _TAP_DISTANCE_THRESHOLD,
|
195 |
+
annotation_width_augment_fraction = ANNOTATION_WIDTH_AUGMENT_FRACTION,
|
196 |
+
annotation_height_augment_fraction = ANNOTATION_HEIGHT_AUGMENT_FRACTION,
|
197 |
+
):
|
198 |
+
"""Determines if two actions are considered to be the same.
|
199 |
+
|
200 |
+
Two actions being "the same" is defined here as two actions that would result
|
201 |
+
in a similar screen state.
|
202 |
+
|
203 |
+
Args:
|
204 |
+
action_1_touch_yx: The (y, x) coordinates of the first action's touch.
|
205 |
+
action_1_lift_yx: The (y, x) coordinates of the first action's lift.
|
206 |
+
action_1_action_type: The action type of the first action.
|
207 |
+
action_2_touch_yx: The (y, x) coordinates of the second action's touch.
|
208 |
+
action_2_lift_yx: The (y, x) coordinates of the second action's lift.
|
209 |
+
action_2_action_type: The action type of the second action.
|
210 |
+
annotation_positions: The positions of the UI annotations for the screen. It
|
211 |
+
is A 2D int array of shape (num_bboxes, 4), where each row represents a
|
212 |
+
bounding box: (y_top_left, x_top_left, box_height, box_width). Note that
|
213 |
+
containment is inclusive of the bounding box edges.
|
214 |
+
tap_distance_threshold: The threshold that determines if two taps result in
|
215 |
+
a matching screen state if they don't fall the same bounding boxes.
|
216 |
+
annotation_width_augment_fraction: The fraction to increase the width of the
|
217 |
+
bounding box by.
|
218 |
+
annotation_height_augment_fraction: The fraction to increase the height of
|
219 |
+
of the bounding box by.
|
220 |
+
|
221 |
+
Returns:
|
222 |
+
A boolean representing whether the two given actions are the same or not.
|
223 |
+
"""
|
224 |
+
action_1_touch_yx = jnp.asarray(action_1_touch_yx)
|
225 |
+
action_1_lift_yx = jnp.asarray(action_1_lift_yx)
|
226 |
+
action_2_touch_yx = jnp.asarray(action_2_touch_yx)
|
227 |
+
action_2_lift_yx = jnp.asarray(action_2_lift_yx)
|
228 |
+
|
229 |
+
# Checks if at least one of the actions is global (i.e. not DUAL_POINT),
|
230 |
+
# because if that is the case, only the actions' types need to be compared.
|
231 |
+
has_non_dual_point_action = jnp.logical_or(
|
232 |
+
_is_non_dual_point_action(action_1_action_type),
|
233 |
+
_is_non_dual_point_action(action_2_action_type),
|
234 |
+
)
|
235 |
+
#print("non dual point: "+str(has_non_dual_point_action))
|
236 |
+
|
237 |
+
different_dual_point_types = jnp.logical_xor(
|
238 |
+
is_tap_action(action_1_touch_yx, action_1_lift_yx),
|
239 |
+
is_tap_action(action_2_touch_yx, action_2_lift_yx),
|
240 |
+
)
|
241 |
+
#print("different dual type: "+str(different_dual_point_types))
|
242 |
+
|
243 |
+
is_tap = jnp.logical_and(
|
244 |
+
is_tap_action(action_1_touch_yx, action_1_lift_yx),
|
245 |
+
is_tap_action(action_2_touch_yx, action_2_lift_yx),
|
246 |
+
)
|
247 |
+
#print("is tap: "+str(is_tap))
|
248 |
+
|
249 |
+
taps_match = _check_tap_actions_match(
|
250 |
+
action_1_touch_yx,
|
251 |
+
action_2_touch_yx,
|
252 |
+
annotation_positions,
|
253 |
+
tap_distance_threshold,
|
254 |
+
annotation_width_augment_fraction,
|
255 |
+
annotation_height_augment_fraction,
|
256 |
+
)
|
257 |
+
#print("tap match: "+str(taps_match))
|
258 |
+
|
259 |
+
taps_match = jnp.logical_and(is_tap, taps_match)
|
260 |
+
#print("tap match: "+str(taps_match))
|
261 |
+
|
262 |
+
drags_match = _check_drag_actions_match(
|
263 |
+
action_1_touch_yx, action_1_lift_yx, action_2_touch_yx, action_2_lift_yx
|
264 |
+
)
|
265 |
+
drags_match = jnp.where(is_tap, False, drags_match)
|
266 |
+
#print("drag match: "+str(drags_match))
|
267 |
+
|
268 |
+
return jnp.where(
|
269 |
+
has_non_dual_point_action,
|
270 |
+
jnp.equal(action_1_action_type, action_2_action_type),
|
271 |
+
jnp.where(
|
272 |
+
different_dual_point_types,
|
273 |
+
False,
|
274 |
+
jnp.logical_or(taps_match, drags_match),
|
275 |
+
),
|
276 |
+
)
|
277 |
+
|
278 |
+
|
279 |
+
def action_2_format(step_data):
|
280 |
+
# 把test数据集中的动作格式转换为计算matching score的格式
|
281 |
+
action_type = step_data["action_type_id"]
|
282 |
+
|
283 |
+
if action_type == 4:
|
284 |
+
if step_data["action_type_text"] == 'click': # 点击
|
285 |
+
touch_point = step_data["touch"]
|
286 |
+
lift_point = step_data["lift"]
|
287 |
+
else: # 上下左右滑动
|
288 |
+
if step_data["action_type_text"] == 'scroll down':
|
289 |
+
touch_point = [0.5, 0.8]
|
290 |
+
lift_point = [0.5, 0.2]
|
291 |
+
elif step_data["action_type_text"] == 'scroll up':
|
292 |
+
touch_point = [0.5, 0.2]
|
293 |
+
lift_point = [0.5, 0.8]
|
294 |
+
elif step_data["action_type_text"] == 'scroll left':
|
295 |
+
touch_point = [0.2, 0.5]
|
296 |
+
lift_point = [0.8, 0.5]
|
297 |
+
elif step_data["action_type_text"] == 'scroll right':
|
298 |
+
touch_point = [0.8, 0.5]
|
299 |
+
lift_point = [0.2, 0.5]
|
300 |
+
else:
|
301 |
+
touch_point = [-1.0, -1.0]
|
302 |
+
lift_point = [-1.0, -1.0]
|
303 |
+
|
304 |
+
if action_type == 3:
|
305 |
+
typed_text = step_data["type_text"]
|
306 |
+
else:
|
307 |
+
typed_text = ""
|
308 |
+
|
309 |
+
action = {"action_type": action_type, "touch_point": touch_point, "lift_point": lift_point,
|
310 |
+
"typed_text": typed_text}
|
311 |
+
|
312 |
+
action["touch_point"] = [action["touch_point"][1], action["touch_point"][0]]
|
313 |
+
action["lift_point"] = [action["lift_point"][1], action["lift_point"][0]]
|
314 |
+
action["typed_text"] = action["typed_text"].lower()
|
315 |
+
|
316 |
+
return action
|
317 |
+
|
318 |
+
|
319 |
+
def pred_2_format(step_data):
|
320 |
+
# 把模型输出的内容转换为计算action_matching的格式
|
321 |
+
action_type = step_data["action_type"]
|
322 |
+
|
323 |
+
if action_type == 4: # 点击
|
324 |
+
action_type_new = 4
|
325 |
+
touch_point = step_data["click_point"]
|
326 |
+
lift_point = step_data["click_point"]
|
327 |
+
typed_text = ""
|
328 |
+
elif action_type == 0:
|
329 |
+
action_type_new = 4
|
330 |
+
touch_point = [0.5, 0.8]
|
331 |
+
lift_point = [0.5, 0.2]
|
332 |
+
typed_text = ""
|
333 |
+
elif action_type == 1:
|
334 |
+
action_type_new = 4
|
335 |
+
touch_point = [0.5, 0.2]
|
336 |
+
lift_point = [0.5, 0.8]
|
337 |
+
typed_text = ""
|
338 |
+
elif action_type == 8:
|
339 |
+
action_type_new = 4
|
340 |
+
touch_point = [0.2, 0.5]
|
341 |
+
lift_point = [0.8, 0.5]
|
342 |
+
typed_text = ""
|
343 |
+
elif action_type == 9:
|
344 |
+
action_type_new = 4
|
345 |
+
touch_point = [0.8, 0.5]
|
346 |
+
lift_point = [0.2, 0.5]
|
347 |
+
typed_text = ""
|
348 |
+
else:
|
349 |
+
action_type_new = action_type
|
350 |
+
touch_point = [-1.0, -1.0]
|
351 |
+
lift_point = [-1.0, -1.0]
|
352 |
+
typed_text = ""
|
353 |
+
if action_type_new == 3:
|
354 |
+
typed_text = step_data["typed_text"]
|
355 |
+
|
356 |
+
action = {"action_type": action_type_new, "touch_point": touch_point, "lift_point": lift_point,
|
357 |
+
"typed_text": typed_text}
|
358 |
+
|
359 |
+
action["touch_point"] = [action["touch_point"][1], action["touch_point"][0]]
|
360 |
+
action["lift_point"] = [action["lift_point"][1], action["lift_point"][0]]
|
361 |
+
action["typed_text"] = action["typed_text"].lower()
|
362 |
+
|
363 |
+
return action
|
364 |
+
|
365 |
+
|
366 |
+
def pred_2_format_simplified(step_data):
|
367 |
+
# 把模型输出的内容转换为计算action_matching的格式
|
368 |
+
action_type = step_data["action_type"]
|
369 |
+
|
370 |
+
if action_type == 'click' : # 点击
|
371 |
+
action_type_new = 4
|
372 |
+
touch_point = step_data["click_point"]
|
373 |
+
lift_point = step_data["click_point"]
|
374 |
+
typed_text = ""
|
375 |
+
elif action_type == 'scroll' and step_data["direction"] == 'down':
|
376 |
+
action_type_new = 4
|
377 |
+
touch_point = [0.5, 0.8]
|
378 |
+
lift_point = [0.5, 0.2]
|
379 |
+
typed_text = ""
|
380 |
+
elif action_type == 'scroll' and step_data["direction"] == 'up':
|
381 |
+
action_type_new = 4
|
382 |
+
touch_point = [0.5, 0.2]
|
383 |
+
lift_point = [0.5, 0.8]
|
384 |
+
typed_text = ""
|
385 |
+
elif action_type == 'scroll' and step_data["direction"] == 'left':
|
386 |
+
action_type_new = 4
|
387 |
+
touch_point = [0.2, 0.5]
|
388 |
+
lift_point = [0.8, 0.5]
|
389 |
+
typed_text = ""
|
390 |
+
elif action_type == 'scroll' and step_data["direction"] == 'right':
|
391 |
+
action_type_new = 4
|
392 |
+
touch_point = [0.8, 0.5]
|
393 |
+
lift_point = [0.2, 0.5]
|
394 |
+
typed_text = ""
|
395 |
+
elif action_type == 'type':
|
396 |
+
action_type_new = 3
|
397 |
+
touch_point = [-1.0, -1.0]
|
398 |
+
lift_point = [-1.0, -1.0]
|
399 |
+
typed_text = step_data["text"]
|
400 |
+
elif action_type == 'navigate_back':
|
401 |
+
action_type_new = 5
|
402 |
+
touch_point = [-1.0, -1.0]
|
403 |
+
lift_point = [-1.0, -1.0]
|
404 |
+
typed_text = ""
|
405 |
+
elif action_type == 'navigate_home':
|
406 |
+
action_type_new = 6
|
407 |
+
touch_point = [-1.0, -1.0]
|
408 |
+
lift_point = [-1.0, -1.0]
|
409 |
+
typed_text = ""
|
410 |
+
else:
|
411 |
+
action_type_new = action_type
|
412 |
+
touch_point = [-1.0, -1.0]
|
413 |
+
lift_point = [-1.0, -1.0]
|
414 |
+
typed_text = ""
|
415 |
+
# if action_type_new == 'type':
|
416 |
+
# typed_text = step_data["text"]
|
417 |
+
|
418 |
+
action = {"action_type": action_type_new, "touch_point": touch_point, "lift_point": lift_point,
|
419 |
+
"typed_text": typed_text}
|
420 |
+
|
421 |
+
action["touch_point"] = [action["touch_point"][1], action["touch_point"][0]]
|
422 |
+
action["lift_point"] = [action["lift_point"][1], action["lift_point"][0]]
|
423 |
+
action["typed_text"] = action["typed_text"].lower()
|
424 |
+
|
425 |
+
return action
|
util/action_type.py
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
'''
|
2 |
+
Adapted from https://github.com/google-research/google-research/tree/master/android_in_the_wild
|
3 |
+
'''
|
4 |
+
|
5 |
+
import enum
|
6 |
+
|
7 |
+
class ActionType(enum.IntEnum):
|
8 |
+
|
9 |
+
# Placeholders for unused enum values
|
10 |
+
UNUSED_0 = 0
|
11 |
+
UNUSED_1 = 1
|
12 |
+
UNUSED_2 = 2
|
13 |
+
UNUSED_8 = 8
|
14 |
+
UNUSED_9 = 9
|
15 |
+
|
16 |
+
########### Agent actions ###########
|
17 |
+
|
18 |
+
# A type action that sends text to the emulator. Note that this simply sends
|
19 |
+
# text and does not perform any clicks for element focus or enter presses for
|
20 |
+
# submitting text.
|
21 |
+
TYPE = 3
|
22 |
+
|
23 |
+
# The dual point action used to represent all gestures.
|
24 |
+
DUAL_POINT = 4
|
25 |
+
|
26 |
+
# These actions differentiate pressing the home and back button from touches.
|
27 |
+
# They represent explicit presses of back and home performed using ADB.
|
28 |
+
PRESS_BACK = 5
|
29 |
+
PRESS_HOME = 6
|
30 |
+
|
31 |
+
# An action representing that ADB command for hitting enter was performed.
|
32 |
+
PRESS_ENTER = 7
|
33 |
+
|
34 |
+
########### Episode status actions ###########
|
35 |
+
|
36 |
+
# An action used to indicate the desired task has been completed and resets
|
37 |
+
# the environment. This action should also be used in the case that the task
|
38 |
+
# has already been completed and there is nothing to do.
|
39 |
+
# e.g. The task is to turn on the Wi-Fi when it is already on
|
40 |
+
STATUS_TASK_COMPLETE = 10
|
41 |
+
|
42 |
+
# An action used to indicate that desired task is impossible to complete and
|
43 |
+
# resets the environment. This can be a result of many different things
|
44 |
+
# including UI changes, Android version differences, etc.
|
45 |
+
STATUS_TASK_IMPOSSIBLE = 11
|
util/box_annotator.py
ADDED
@@ -0,0 +1,262 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import List, Optional, Union, Tuple
|
2 |
+
|
3 |
+
import cv2
|
4 |
+
import numpy as np
|
5 |
+
|
6 |
+
from supervision.detection.core import Detections
|
7 |
+
from supervision.draw.color import Color, ColorPalette
|
8 |
+
|
9 |
+
|
10 |
+
class BoxAnnotator:
|
11 |
+
"""
|
12 |
+
A class for drawing bounding boxes on an image using detections provided.
|
13 |
+
|
14 |
+
Attributes:
|
15 |
+
color (Union[Color, ColorPalette]): The color to draw the bounding box,
|
16 |
+
can be a single color or a color palette
|
17 |
+
thickness (int): The thickness of the bounding box lines, default is 2
|
18 |
+
text_color (Color): The color of the text on the bounding box, default is white
|
19 |
+
text_scale (float): The scale of the text on the bounding box, default is 0.5
|
20 |
+
text_thickness (int): The thickness of the text on the bounding box,
|
21 |
+
default is 1
|
22 |
+
text_padding (int): The padding around the text on the bounding box,
|
23 |
+
default is 5
|
24 |
+
|
25 |
+
"""
|
26 |
+
|
27 |
+
def __init__(
|
28 |
+
self,
|
29 |
+
color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
|
30 |
+
thickness: int = 3, # 1 for seeclick 2 for mind2web and 3 for demo
|
31 |
+
text_color: Color = Color.BLACK,
|
32 |
+
text_scale: float = 0.5, # 0.8 for mobile/web, 0.3 for desktop # 0.4 for mind2web
|
33 |
+
text_thickness: int = 2, #1, # 2 for demo
|
34 |
+
text_padding: int = 10,
|
35 |
+
avoid_overlap: bool = True,
|
36 |
+
):
|
37 |
+
self.color: Union[Color, ColorPalette] = color
|
38 |
+
self.thickness: int = thickness
|
39 |
+
self.text_color: Color = text_color
|
40 |
+
self.text_scale: float = text_scale
|
41 |
+
self.text_thickness: int = text_thickness
|
42 |
+
self.text_padding: int = text_padding
|
43 |
+
self.avoid_overlap: bool = avoid_overlap
|
44 |
+
|
45 |
+
def annotate(
|
46 |
+
self,
|
47 |
+
scene: np.ndarray,
|
48 |
+
detections: Detections,
|
49 |
+
labels: Optional[List[str]] = None,
|
50 |
+
skip_label: bool = False,
|
51 |
+
image_size: Optional[Tuple[int, int]] = None,
|
52 |
+
) -> np.ndarray:
|
53 |
+
"""
|
54 |
+
Draws bounding boxes on the frame using the detections provided.
|
55 |
+
|
56 |
+
Args:
|
57 |
+
scene (np.ndarray): The image on which the bounding boxes will be drawn
|
58 |
+
detections (Detections): The detections for which the
|
59 |
+
bounding boxes will be drawn
|
60 |
+
labels (Optional[List[str]]): An optional list of labels
|
61 |
+
corresponding to each detection. If `labels` are not provided,
|
62 |
+
corresponding `class_id` will be used as label.
|
63 |
+
skip_label (bool): Is set to `True`, skips bounding box label annotation.
|
64 |
+
Returns:
|
65 |
+
np.ndarray: The image with the bounding boxes drawn on it
|
66 |
+
|
67 |
+
Example:
|
68 |
+
```python
|
69 |
+
import supervision as sv
|
70 |
+
|
71 |
+
classes = ['person', ...]
|
72 |
+
image = ...
|
73 |
+
detections = sv.Detections(...)
|
74 |
+
|
75 |
+
box_annotator = sv.BoxAnnotator()
|
76 |
+
labels = [
|
77 |
+
f"{classes[class_id]} {confidence:0.2f}"
|
78 |
+
for _, _, confidence, class_id, _ in detections
|
79 |
+
]
|
80 |
+
annotated_frame = box_annotator.annotate(
|
81 |
+
scene=image.copy(),
|
82 |
+
detections=detections,
|
83 |
+
labels=labels
|
84 |
+
)
|
85 |
+
```
|
86 |
+
"""
|
87 |
+
font = cv2.FONT_HERSHEY_SIMPLEX
|
88 |
+
for i in range(len(detections)):
|
89 |
+
x1, y1, x2, y2 = detections.xyxy[i].astype(int)
|
90 |
+
class_id = (
|
91 |
+
detections.class_id[i] if detections.class_id is not None else None
|
92 |
+
)
|
93 |
+
idx = class_id if class_id is not None else i
|
94 |
+
color = (
|
95 |
+
self.color.by_idx(idx)
|
96 |
+
if isinstance(self.color, ColorPalette)
|
97 |
+
else self.color
|
98 |
+
)
|
99 |
+
cv2.rectangle(
|
100 |
+
img=scene,
|
101 |
+
pt1=(x1, y1),
|
102 |
+
pt2=(x2, y2),
|
103 |
+
color=color.as_bgr(),
|
104 |
+
thickness=self.thickness,
|
105 |
+
)
|
106 |
+
if skip_label:
|
107 |
+
continue
|
108 |
+
|
109 |
+
text = (
|
110 |
+
f"{class_id}"
|
111 |
+
if (labels is None or len(detections) != len(labels))
|
112 |
+
else labels[i]
|
113 |
+
)
|
114 |
+
|
115 |
+
text_width, text_height = cv2.getTextSize(
|
116 |
+
text=text,
|
117 |
+
fontFace=font,
|
118 |
+
fontScale=self.text_scale,
|
119 |
+
thickness=self.text_thickness,
|
120 |
+
)[0]
|
121 |
+
|
122 |
+
if not self.avoid_overlap:
|
123 |
+
text_x = x1 + self.text_padding
|
124 |
+
text_y = y1 - self.text_padding
|
125 |
+
|
126 |
+
text_background_x1 = x1
|
127 |
+
text_background_y1 = y1 - 2 * self.text_padding - text_height
|
128 |
+
|
129 |
+
text_background_x2 = x1 + 2 * self.text_padding + text_width
|
130 |
+
text_background_y2 = y1
|
131 |
+
# text_x = x1 - self.text_padding - text_width
|
132 |
+
# text_y = y1 + self.text_padding + text_height
|
133 |
+
# text_background_x1 = x1 - 2 * self.text_padding - text_width
|
134 |
+
# text_background_y1 = y1
|
135 |
+
# text_background_x2 = x1
|
136 |
+
# text_background_y2 = y1 + 2 * self.text_padding + text_height
|
137 |
+
else:
|
138 |
+
text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2 = get_optimal_label_pos(self.text_padding, text_width, text_height, x1, y1, x2, y2, detections, image_size)
|
139 |
+
|
140 |
+
cv2.rectangle(
|
141 |
+
img=scene,
|
142 |
+
pt1=(text_background_x1, text_background_y1),
|
143 |
+
pt2=(text_background_x2, text_background_y2),
|
144 |
+
color=color.as_bgr(),
|
145 |
+
thickness=cv2.FILLED,
|
146 |
+
)
|
147 |
+
# import pdb; pdb.set_trace()
|
148 |
+
box_color = color.as_rgb()
|
149 |
+
luminance = 0.299 * box_color[0] + 0.587 * box_color[1] + 0.114 * box_color[2]
|
150 |
+
text_color = (0,0,0) if luminance > 160 else (255,255,255)
|
151 |
+
cv2.putText(
|
152 |
+
img=scene,
|
153 |
+
text=text,
|
154 |
+
org=(text_x, text_y),
|
155 |
+
fontFace=font,
|
156 |
+
fontScale=self.text_scale,
|
157 |
+
# color=self.text_color.as_rgb(),
|
158 |
+
color=text_color,
|
159 |
+
thickness=self.text_thickness,
|
160 |
+
lineType=cv2.LINE_AA,
|
161 |
+
)
|
162 |
+
return scene
|
163 |
+
|
164 |
+
|
165 |
+
def box_area(box):
|
166 |
+
return (box[2] - box[0]) * (box[3] - box[1])
|
167 |
+
|
168 |
+
def intersection_area(box1, box2):
|
169 |
+
x1 = max(box1[0], box2[0])
|
170 |
+
y1 = max(box1[1], box2[1])
|
171 |
+
x2 = min(box1[2], box2[2])
|
172 |
+
y2 = min(box1[3], box2[3])
|
173 |
+
return max(0, x2 - x1) * max(0, y2 - y1)
|
174 |
+
|
175 |
+
def IoU(box1, box2, return_max=True):
|
176 |
+
intersection = intersection_area(box1, box2)
|
177 |
+
union = box_area(box1) + box_area(box2) - intersection
|
178 |
+
if box_area(box1) > 0 and box_area(box2) > 0:
|
179 |
+
ratio1 = intersection / box_area(box1)
|
180 |
+
ratio2 = intersection / box_area(box2)
|
181 |
+
else:
|
182 |
+
ratio1, ratio2 = 0, 0
|
183 |
+
if return_max:
|
184 |
+
return max(intersection / union, ratio1, ratio2)
|
185 |
+
else:
|
186 |
+
return intersection / union
|
187 |
+
|
188 |
+
|
189 |
+
def get_optimal_label_pos(text_padding, text_width, text_height, x1, y1, x2, y2, detections, image_size):
|
190 |
+
""" check overlap of text and background detection box, and get_optimal_label_pos,
|
191 |
+
pos: str, position of the text, must be one of 'top left', 'top right', 'outer left', 'outer right' TODO: if all are overlapping, return the last one, i.e. outer right
|
192 |
+
Threshold: default to 0.3
|
193 |
+
"""
|
194 |
+
|
195 |
+
def get_is_overlap(detections, text_background_x1, text_background_y1, text_background_x2, text_background_y2, image_size):
|
196 |
+
is_overlap = False
|
197 |
+
for i in range(len(detections)):
|
198 |
+
detection = detections.xyxy[i].astype(int)
|
199 |
+
if IoU([text_background_x1, text_background_y1, text_background_x2, text_background_y2], detection) > 0.3:
|
200 |
+
is_overlap = True
|
201 |
+
break
|
202 |
+
# check if the text is out of the image
|
203 |
+
if text_background_x1 < 0 or text_background_x2 > image_size[0] or text_background_y1 < 0 or text_background_y2 > image_size[1]:
|
204 |
+
is_overlap = True
|
205 |
+
return is_overlap
|
206 |
+
|
207 |
+
# if pos == 'top left':
|
208 |
+
text_x = x1 + text_padding
|
209 |
+
text_y = y1 - text_padding
|
210 |
+
|
211 |
+
text_background_x1 = x1
|
212 |
+
text_background_y1 = y1 - 2 * text_padding - text_height
|
213 |
+
|
214 |
+
text_background_x2 = x1 + 2 * text_padding + text_width
|
215 |
+
text_background_y2 = y1
|
216 |
+
is_overlap = get_is_overlap(detections, text_background_x1, text_background_y1, text_background_x2, text_background_y2, image_size)
|
217 |
+
if not is_overlap:
|
218 |
+
return text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2
|
219 |
+
|
220 |
+
# elif pos == 'outer left':
|
221 |
+
text_x = x1 - text_padding - text_width
|
222 |
+
text_y = y1 + text_padding + text_height
|
223 |
+
|
224 |
+
text_background_x1 = x1 - 2 * text_padding - text_width
|
225 |
+
text_background_y1 = y1
|
226 |
+
|
227 |
+
text_background_x2 = x1
|
228 |
+
text_background_y2 = y1 + 2 * text_padding + text_height
|
229 |
+
is_overlap = get_is_overlap(detections, text_background_x1, text_background_y1, text_background_x2, text_background_y2, image_size)
|
230 |
+
if not is_overlap:
|
231 |
+
return text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2
|
232 |
+
|
233 |
+
|
234 |
+
# elif pos == 'outer right':
|
235 |
+
text_x = x2 + text_padding
|
236 |
+
text_y = y1 + text_padding + text_height
|
237 |
+
|
238 |
+
text_background_x1 = x2
|
239 |
+
text_background_y1 = y1
|
240 |
+
|
241 |
+
text_background_x2 = x2 + 2 * text_padding + text_width
|
242 |
+
text_background_y2 = y1 + 2 * text_padding + text_height
|
243 |
+
|
244 |
+
is_overlap = get_is_overlap(detections, text_background_x1, text_background_y1, text_background_x2, text_background_y2, image_size)
|
245 |
+
if not is_overlap:
|
246 |
+
return text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2
|
247 |
+
|
248 |
+
# elif pos == 'top right':
|
249 |
+
text_x = x2 - text_padding - text_width
|
250 |
+
text_y = y1 - text_padding
|
251 |
+
|
252 |
+
text_background_x1 = x2 - 2 * text_padding - text_width
|
253 |
+
text_background_y1 = y1 - 2 * text_padding - text_height
|
254 |
+
|
255 |
+
text_background_x2 = x2
|
256 |
+
text_background_y2 = y1
|
257 |
+
|
258 |
+
is_overlap = get_is_overlap(detections, text_background_x1, text_background_y1, text_background_x2, text_background_y2, image_size)
|
259 |
+
if not is_overlap:
|
260 |
+
return text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2
|
261 |
+
|
262 |
+
return text_x, text_y, text_background_x1, text_background_y1, text_background_x2, text_background_y2
|
utils.py
ADDED
@@ -0,0 +1,403 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# from ultralytics import YOLO
|
2 |
+
import os
|
3 |
+
import io
|
4 |
+
import base64
|
5 |
+
import time
|
6 |
+
from PIL import Image, ImageDraw, ImageFont
|
7 |
+
import json
|
8 |
+
import requests
|
9 |
+
# utility function
|
10 |
+
import os
|
11 |
+
from openai import AzureOpenAI
|
12 |
+
|
13 |
+
import json
|
14 |
+
import sys
|
15 |
+
import os
|
16 |
+
import cv2
|
17 |
+
import numpy as np
|
18 |
+
# %matplotlib inline
|
19 |
+
from matplotlib import pyplot as plt
|
20 |
+
import easyocr
|
21 |
+
reader = easyocr.Reader(['en'])
|
22 |
+
import time
|
23 |
+
import base64
|
24 |
+
|
25 |
+
import os
|
26 |
+
import ast
|
27 |
+
import torch
|
28 |
+
from typing import Tuple, List
|
29 |
+
from torchvision.ops import box_convert
|
30 |
+
import re
|
31 |
+
from torchvision.transforms import ToPILImage
|
32 |
+
import supervision as sv
|
33 |
+
import torchvision.transforms as T
|
34 |
+
|
35 |
+
|
36 |
+
def get_caption_model_processor(model_name, model_name_or_path="Salesforce/blip2-opt-2.7b", device=None):
|
37 |
+
if not device:
|
38 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
39 |
+
if model_name == "blip2":
|
40 |
+
from transformers import Blip2Processor, Blip2ForConditionalGeneration
|
41 |
+
processor = Blip2Processor.from_pretrained("Salesforce/blip2-opt-2.7b")
|
42 |
+
if device == 'cpu':
|
43 |
+
model = Blip2ForConditionalGeneration.from_pretrained(
|
44 |
+
model_name_or_path, device_map=None, torch_dtype=torch.float32
|
45 |
+
)
|
46 |
+
else:
|
47 |
+
model = Blip2ForConditionalGeneration.from_pretrained(
|
48 |
+
model_name_or_path, device_map=None, torch_dtype=torch.float16
|
49 |
+
).to(device)
|
50 |
+
elif model_name == "florence2":
|
51 |
+
from transformers import AutoProcessor, AutoModelForCausalLM
|
52 |
+
processor = AutoProcessor.from_pretrained("microsoft/Florence-2-base", trust_remote_code=True)
|
53 |
+
if device == 'cpu':
|
54 |
+
model = AutoModelForCausalLM.from_pretrained(model_name_or_path, torch_dtype=torch.float32, trust_remote_code=True)
|
55 |
+
else:
|
56 |
+
model = AutoModelForCausalLM.from_pretrained(model_name_or_path, torch_dtype=torch.float16, trust_remote_code=True).to(device)
|
57 |
+
return {'model': model.to(device), 'processor': processor}
|
58 |
+
|
59 |
+
|
60 |
+
def get_yolo_model(model_path):
|
61 |
+
from ultralytics import YOLO
|
62 |
+
# Load the model.
|
63 |
+
model = YOLO(model_path)
|
64 |
+
return model
|
65 |
+
|
66 |
+
|
67 |
+
@torch.inference_mode()
|
68 |
+
def get_parsed_content_icon(filtered_boxes, ocr_bbox, image_source, caption_model_processor, prompt=None):
|
69 |
+
to_pil = ToPILImage()
|
70 |
+
if ocr_bbox:
|
71 |
+
non_ocr_boxes = filtered_boxes[len(ocr_bbox):]
|
72 |
+
else:
|
73 |
+
non_ocr_boxes = filtered_boxes
|
74 |
+
croped_pil_image = []
|
75 |
+
for i, coord in enumerate(non_ocr_boxes):
|
76 |
+
xmin, xmax = int(coord[0]*image_source.shape[1]), int(coord[2]*image_source.shape[1])
|
77 |
+
ymin, ymax = int(coord[1]*image_source.shape[0]), int(coord[3]*image_source.shape[0])
|
78 |
+
cropped_image = image_source[ymin:ymax, xmin:xmax, :]
|
79 |
+
croped_pil_image.append(to_pil(cropped_image))
|
80 |
+
|
81 |
+
model, processor = caption_model_processor['model'], caption_model_processor['processor']
|
82 |
+
if not prompt:
|
83 |
+
if 'florence' in model.config.name_or_path:
|
84 |
+
prompt = "<CAPTION>"
|
85 |
+
else:
|
86 |
+
prompt = "The image shows"
|
87 |
+
|
88 |
+
batch_size = 10 # Number of samples per batch
|
89 |
+
generated_texts = []
|
90 |
+
device = model.device
|
91 |
+
|
92 |
+
for i in range(0, len(croped_pil_image), batch_size):
|
93 |
+
batch = croped_pil_image[i:i+batch_size]
|
94 |
+
if model.device.type == 'cuda':
|
95 |
+
inputs = processor(images=batch, text=[prompt]*len(batch), return_tensors="pt").to(device=device, dtype=torch.float16)
|
96 |
+
else:
|
97 |
+
inputs = processor(images=batch, text=[prompt]*len(batch), return_tensors="pt").to(device=device)
|
98 |
+
if 'florence' in model.config.name_or_path:
|
99 |
+
generated_ids = model.generate(input_ids=inputs["input_ids"],pixel_values=inputs["pixel_values"],max_new_tokens=1024,num_beams=3, do_sample=False)
|
100 |
+
else:
|
101 |
+
generated_ids = model.generate(**inputs, max_length=100, num_beams=5, no_repeat_ngram_size=2, early_stopping=True, num_return_sequences=1) # temperature=0.01, do_sample=True,
|
102 |
+
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)
|
103 |
+
generated_text = [gen.strip() for gen in generated_text]
|
104 |
+
generated_texts.extend(generated_text)
|
105 |
+
|
106 |
+
return generated_texts
|
107 |
+
|
108 |
+
|
109 |
+
|
110 |
+
def get_parsed_content_icon_phi3v(filtered_boxes, ocr_bbox, image_source, caption_model_processor):
|
111 |
+
to_pil = ToPILImage()
|
112 |
+
if ocr_bbox:
|
113 |
+
non_ocr_boxes = filtered_boxes[len(ocr_bbox):]
|
114 |
+
else:
|
115 |
+
non_ocr_boxes = filtered_boxes
|
116 |
+
croped_pil_image = []
|
117 |
+
for i, coord in enumerate(non_ocr_boxes):
|
118 |
+
xmin, xmax = int(coord[0]*image_source.shape[1]), int(coord[2]*image_source.shape[1])
|
119 |
+
ymin, ymax = int(coord[1]*image_source.shape[0]), int(coord[3]*image_source.shape[0])
|
120 |
+
cropped_image = image_source[ymin:ymax, xmin:xmax, :]
|
121 |
+
croped_pil_image.append(to_pil(cropped_image))
|
122 |
+
|
123 |
+
model, processor = caption_model_processor['model'], caption_model_processor['processor']
|
124 |
+
device = model.device
|
125 |
+
messages = [{"role": "user", "content": "<|image_1|>\ndescribe the icon in one sentence"}]
|
126 |
+
prompt = processor.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
127 |
+
|
128 |
+
batch_size = 5 # Number of samples per batch
|
129 |
+
generated_texts = []
|
130 |
+
|
131 |
+
for i in range(0, len(croped_pil_image), batch_size):
|
132 |
+
images = croped_pil_image[i:i+batch_size]
|
133 |
+
image_inputs = [processor.image_processor(x, return_tensors="pt") for x in images]
|
134 |
+
inputs ={'input_ids': [], 'attention_mask': [], 'pixel_values': [], 'image_sizes': []}
|
135 |
+
texts = [prompt] * len(images)
|
136 |
+
for i, txt in enumerate(texts):
|
137 |
+
input = processor._convert_images_texts_to_inputs(image_inputs[i], txt, return_tensors="pt")
|
138 |
+
inputs['input_ids'].append(input['input_ids'])
|
139 |
+
inputs['attention_mask'].append(input['attention_mask'])
|
140 |
+
inputs['pixel_values'].append(input['pixel_values'])
|
141 |
+
inputs['image_sizes'].append(input['image_sizes'])
|
142 |
+
max_len = max([x.shape[1] for x in inputs['input_ids']])
|
143 |
+
for i, v in enumerate(inputs['input_ids']):
|
144 |
+
inputs['input_ids'][i] = torch.cat([processor.tokenizer.pad_token_id * torch.ones(1, max_len - v.shape[1], dtype=torch.long), v], dim=1)
|
145 |
+
inputs['attention_mask'][i] = torch.cat([torch.zeros(1, max_len - v.shape[1], dtype=torch.long), inputs['attention_mask'][i]], dim=1)
|
146 |
+
inputs_cat = {k: torch.concatenate(v).to(device) for k, v in inputs.items()}
|
147 |
+
|
148 |
+
generation_args = {
|
149 |
+
"max_new_tokens": 25,
|
150 |
+
"temperature": 0.01,
|
151 |
+
"do_sample": False,
|
152 |
+
}
|
153 |
+
generate_ids = model.generate(**inputs_cat, eos_token_id=processor.tokenizer.eos_token_id, **generation_args)
|
154 |
+
# # remove input tokens
|
155 |
+
generate_ids = generate_ids[:, inputs_cat['input_ids'].shape[1]:]
|
156 |
+
response = processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)
|
157 |
+
response = [res.strip('\n').strip() for res in response]
|
158 |
+
generated_texts.extend(response)
|
159 |
+
|
160 |
+
return generated_texts
|
161 |
+
|
162 |
+
def remove_overlap(boxes, iou_threshold, ocr_bbox=None):
|
163 |
+
assert ocr_bbox is None or isinstance(ocr_bbox, List)
|
164 |
+
|
165 |
+
def box_area(box):
|
166 |
+
return (box[2] - box[0]) * (box[3] - box[1])
|
167 |
+
|
168 |
+
def intersection_area(box1, box2):
|
169 |
+
x1 = max(box1[0], box2[0])
|
170 |
+
y1 = max(box1[1], box2[1])
|
171 |
+
x2 = min(box1[2], box2[2])
|
172 |
+
y2 = min(box1[3], box2[3])
|
173 |
+
return max(0, x2 - x1) * max(0, y2 - y1)
|
174 |
+
|
175 |
+
def IoU(box1, box2):
|
176 |
+
intersection = intersection_area(box1, box2)
|
177 |
+
union = box_area(box1) + box_area(box2) - intersection + 1e-6
|
178 |
+
if box_area(box1) > 0 and box_area(box2) > 0:
|
179 |
+
ratio1 = intersection / box_area(box1)
|
180 |
+
ratio2 = intersection / box_area(box2)
|
181 |
+
else:
|
182 |
+
ratio1, ratio2 = 0, 0
|
183 |
+
return max(intersection / union, ratio1, ratio2)
|
184 |
+
|
185 |
+
boxes = boxes.tolist()
|
186 |
+
filtered_boxes = []
|
187 |
+
if ocr_bbox:
|
188 |
+
filtered_boxes.extend(ocr_bbox)
|
189 |
+
# print('ocr_bbox!!!', ocr_bbox)
|
190 |
+
for i, box1 in enumerate(boxes):
|
191 |
+
# if not any(IoU(box1, box2) > iou_threshold and box_area(box1) > box_area(box2) for j, box2 in enumerate(boxes) if i != j):
|
192 |
+
is_valid_box = True
|
193 |
+
for j, box2 in enumerate(boxes):
|
194 |
+
if i != j and IoU(box1, box2) > iou_threshold and box_area(box1) > box_area(box2):
|
195 |
+
is_valid_box = False
|
196 |
+
break
|
197 |
+
if is_valid_box:
|
198 |
+
# add the following 2 lines to include ocr bbox
|
199 |
+
if ocr_bbox:
|
200 |
+
if not any(IoU(box1, box3) > iou_threshold for k, box3 in enumerate(ocr_bbox)):
|
201 |
+
filtered_boxes.append(box1)
|
202 |
+
else:
|
203 |
+
filtered_boxes.append(box1)
|
204 |
+
return torch.tensor(filtered_boxes)
|
205 |
+
|
206 |
+
def load_image(image_path: str) -> Tuple[np.array, torch.Tensor]:
|
207 |
+
transform = T.Compose(
|
208 |
+
[
|
209 |
+
T.RandomResize([800], max_size=1333),
|
210 |
+
T.ToTensor(),
|
211 |
+
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
|
212 |
+
]
|
213 |
+
)
|
214 |
+
image_source = Image.open(image_path).convert("RGB")
|
215 |
+
image = np.asarray(image_source)
|
216 |
+
image_transformed, _ = transform(image_source, None)
|
217 |
+
return image, image_transformed
|
218 |
+
|
219 |
+
|
220 |
+
def annotate(image_source: np.ndarray, boxes: torch.Tensor, logits: torch.Tensor, phrases: List[str], text_scale: float,
|
221 |
+
text_padding=5, text_thickness=2, thickness=3) -> np.ndarray:
|
222 |
+
"""
|
223 |
+
This function annotates an image with bounding boxes and labels.
|
224 |
+
|
225 |
+
Parameters:
|
226 |
+
image_source (np.ndarray): The source image to be annotated.
|
227 |
+
boxes (torch.Tensor): A tensor containing bounding box coordinates. in cxcywh format, pixel scale
|
228 |
+
logits (torch.Tensor): A tensor containing confidence scores for each bounding box.
|
229 |
+
phrases (List[str]): A list of labels for each bounding box.
|
230 |
+
text_scale (float): The scale of the text to be displayed. 0.8 for mobile/web, 0.3 for desktop # 0.4 for mind2web
|
231 |
+
|
232 |
+
Returns:
|
233 |
+
np.ndarray: The annotated image.
|
234 |
+
"""
|
235 |
+
h, w, _ = image_source.shape
|
236 |
+
boxes = boxes * torch.Tensor([w, h, w, h])
|
237 |
+
xyxy = box_convert(boxes=boxes, in_fmt="cxcywh", out_fmt="xyxy").numpy()
|
238 |
+
xywh = box_convert(boxes=boxes, in_fmt="cxcywh", out_fmt="xywh").numpy()
|
239 |
+
detections = sv.Detections(xyxy=xyxy)
|
240 |
+
|
241 |
+
labels = [f"{phrase}" for phrase in range(boxes.shape[0])]
|
242 |
+
|
243 |
+
from util.box_annotator import BoxAnnotator
|
244 |
+
box_annotator = BoxAnnotator(text_scale=text_scale, text_padding=text_padding,text_thickness=text_thickness,thickness=thickness) # 0.8 for mobile/web, 0.3 for desktop # 0.4 for mind2web
|
245 |
+
annotated_frame = image_source.copy()
|
246 |
+
annotated_frame = box_annotator.annotate(scene=annotated_frame, detections=detections, labels=labels, image_size=(w,h))
|
247 |
+
|
248 |
+
label_coordinates = {f"{phrase}": v for phrase, v in zip(phrases, xywh)}
|
249 |
+
return annotated_frame, label_coordinates
|
250 |
+
|
251 |
+
|
252 |
+
def predict(model, image, caption, box_threshold, text_threshold):
|
253 |
+
""" Use huggingface model to replace the original model
|
254 |
+
"""
|
255 |
+
model, processor = model['model'], model['processor']
|
256 |
+
device = model.device
|
257 |
+
|
258 |
+
inputs = processor(images=image, text=caption, return_tensors="pt").to(device)
|
259 |
+
with torch.no_grad():
|
260 |
+
outputs = model(**inputs)
|
261 |
+
|
262 |
+
results = processor.post_process_grounded_object_detection(
|
263 |
+
outputs,
|
264 |
+
inputs.input_ids,
|
265 |
+
box_threshold=box_threshold, # 0.4,
|
266 |
+
text_threshold=text_threshold, # 0.3,
|
267 |
+
target_sizes=[image.size[::-1]]
|
268 |
+
)[0]
|
269 |
+
boxes, logits, phrases = results["boxes"], results["scores"], results["labels"]
|
270 |
+
return boxes, logits, phrases
|
271 |
+
|
272 |
+
|
273 |
+
def predict_yolo(model, image_path, box_threshold):
|
274 |
+
""" Use huggingface model to replace the original model
|
275 |
+
"""
|
276 |
+
# model = model['model']
|
277 |
+
|
278 |
+
result = model.predict(
|
279 |
+
source=image_path,
|
280 |
+
conf=box_threshold,
|
281 |
+
# iou=0.5, # default 0.7
|
282 |
+
)
|
283 |
+
boxes = result[0].boxes.xyxy#.tolist() # in pixel space
|
284 |
+
conf = result[0].boxes.conf
|
285 |
+
phrases = [str(i) for i in range(len(boxes))]
|
286 |
+
|
287 |
+
return boxes, conf, phrases
|
288 |
+
|
289 |
+
|
290 |
+
def get_som_labeled_img(img_path, model=None, BOX_TRESHOLD = 0.01, output_coord_in_ratio=False, ocr_bbox=None, text_scale=0.4, text_padding=5, draw_bbox_config=None, caption_model_processor=None, ocr_text=[], use_local_semantics=True, iou_threshold=0.9,prompt=None):
|
291 |
+
""" ocr_bbox: list of xyxy format bbox
|
292 |
+
"""
|
293 |
+
TEXT_PROMPT = "clickable buttons on the screen"
|
294 |
+
# BOX_TRESHOLD = 0.02 # 0.05/0.02 for web and 0.1 for mobile
|
295 |
+
TEXT_TRESHOLD = 0.01 # 0.9 # 0.01
|
296 |
+
image_source = Image.open(img_path).convert("RGB")
|
297 |
+
w, h = image_source.size
|
298 |
+
# import pdb; pdb.set_trace()
|
299 |
+
if False: # TODO
|
300 |
+
xyxy, logits, phrases = predict(model=model, image=image_source, caption=TEXT_PROMPT, box_threshold=BOX_TRESHOLD, text_threshold=TEXT_TRESHOLD)
|
301 |
+
else:
|
302 |
+
xyxy, logits, phrases = predict_yolo(model=model, image_path=img_path, box_threshold=BOX_TRESHOLD)
|
303 |
+
xyxy = xyxy / torch.Tensor([w, h, w, h]).to(xyxy.device)
|
304 |
+
image_source = np.asarray(image_source)
|
305 |
+
phrases = [str(i) for i in range(len(phrases))]
|
306 |
+
|
307 |
+
# annotate the image with labels
|
308 |
+
h, w, _ = image_source.shape
|
309 |
+
if ocr_bbox:
|
310 |
+
ocr_bbox = torch.tensor(ocr_bbox) / torch.Tensor([w, h, w, h])
|
311 |
+
ocr_bbox=ocr_bbox.tolist()
|
312 |
+
else:
|
313 |
+
print('no ocr bbox!!!')
|
314 |
+
ocr_bbox = None
|
315 |
+
filtered_boxes = remove_overlap(boxes=xyxy, iou_threshold=iou_threshold, ocr_bbox=ocr_bbox)
|
316 |
+
|
317 |
+
# get parsed icon local semantics
|
318 |
+
if use_local_semantics:
|
319 |
+
caption_model = caption_model_processor['model']
|
320 |
+
if 'phi3_v' in caption_model.config.model_type:
|
321 |
+
parsed_content_icon = get_parsed_content_icon_phi3v(filtered_boxes, ocr_bbox, image_source, caption_model_processor)
|
322 |
+
else:
|
323 |
+
parsed_content_icon = get_parsed_content_icon(filtered_boxes, ocr_bbox, image_source, caption_model_processor, prompt=prompt)
|
324 |
+
ocr_text = [f"Text Box ID {i}: {txt}" for i, txt in enumerate(ocr_text)]
|
325 |
+
icon_start = len(ocr_text)
|
326 |
+
parsed_content_icon_ls = []
|
327 |
+
for i, txt in enumerate(parsed_content_icon):
|
328 |
+
parsed_content_icon_ls.append(f"Icon Box ID {str(i+icon_start)}: {txt}")
|
329 |
+
parsed_content_merged = ocr_text + parsed_content_icon_ls
|
330 |
+
else:
|
331 |
+
ocr_text = [f"Text Box ID {i}: {txt}" for i, txt in enumerate(ocr_text)]
|
332 |
+
parsed_content_merged = ocr_text
|
333 |
+
|
334 |
+
filtered_boxes = box_convert(boxes=filtered_boxes, in_fmt="xyxy", out_fmt="cxcywh")
|
335 |
+
|
336 |
+
phrases = [i for i in range(len(filtered_boxes))]
|
337 |
+
|
338 |
+
# draw boxes
|
339 |
+
if draw_bbox_config:
|
340 |
+
annotated_frame, label_coordinates = annotate(image_source=image_source, boxes=filtered_boxes, logits=logits, phrases=phrases, **draw_bbox_config)
|
341 |
+
else:
|
342 |
+
annotated_frame, label_coordinates = annotate(image_source=image_source, boxes=filtered_boxes, logits=logits, phrases=phrases, text_scale=text_scale, text_padding=text_padding)
|
343 |
+
|
344 |
+
pil_img = Image.fromarray(annotated_frame)
|
345 |
+
buffered = io.BytesIO()
|
346 |
+
pil_img.save(buffered, format="PNG")
|
347 |
+
encoded_image = base64.b64encode(buffered.getvalue()).decode('ascii')
|
348 |
+
if output_coord_in_ratio:
|
349 |
+
# h, w, _ = image_source.shape
|
350 |
+
label_coordinates = {k: [v[0]/w, v[1]/h, v[2]/w, v[3]/h] for k, v in label_coordinates.items()}
|
351 |
+
assert w == annotated_frame.shape[1] and h == annotated_frame.shape[0]
|
352 |
+
|
353 |
+
return encoded_image, label_coordinates, parsed_content_merged
|
354 |
+
|
355 |
+
|
356 |
+
def get_xywh(input):
|
357 |
+
x, y, w, h = input[0][0], input[0][1], input[2][0] - input[0][0], input[2][1] - input[0][1]
|
358 |
+
x, y, w, h = int(x), int(y), int(w), int(h)
|
359 |
+
return x, y, w, h
|
360 |
+
|
361 |
+
def get_xyxy(input):
|
362 |
+
x, y, xp, yp = input[0][0], input[0][1], input[2][0], input[2][1]
|
363 |
+
x, y, xp, yp = int(x), int(y), int(xp), int(yp)
|
364 |
+
return x, y, xp, yp
|
365 |
+
|
366 |
+
def get_xywh_yolo(input):
|
367 |
+
x, y, w, h = input[0], input[1], input[2] - input[0], input[3] - input[1]
|
368 |
+
x, y, w, h = int(x), int(y), int(w), int(h)
|
369 |
+
return x, y, w, h
|
370 |
+
|
371 |
+
|
372 |
+
|
373 |
+
def check_ocr_box(image_path, display_img = True, output_bb_format='xywh', goal_filtering=None, easyocr_args=None):
|
374 |
+
if easyocr_args is None:
|
375 |
+
easyocr_args = {}
|
376 |
+
result = reader.readtext(image_path, **easyocr_args)
|
377 |
+
is_goal_filtered = False
|
378 |
+
# print('goal filtering pred:', result[-5:])
|
379 |
+
coord = [item[0] for item in result]
|
380 |
+
text = [item[1] for item in result]
|
381 |
+
# read the image using cv2
|
382 |
+
if display_img:
|
383 |
+
opencv_img = cv2.imread(image_path)
|
384 |
+
opencv_img = cv2.cvtColor(opencv_img, cv2.COLOR_RGB2BGR)
|
385 |
+
bb = []
|
386 |
+
for item in coord:
|
387 |
+
x, y, a, b = get_xywh(item)
|
388 |
+
# print(x, y, a, b)
|
389 |
+
bb.append((x, y, a, b))
|
390 |
+
cv2.rectangle(opencv_img, (x, y), (x+a, y+b), (0, 255, 0), 2)
|
391 |
+
|
392 |
+
# Display the image
|
393 |
+
plt.imshow(opencv_img)
|
394 |
+
else:
|
395 |
+
if output_bb_format == 'xywh':
|
396 |
+
bb = [get_xywh(item) for item in coord]
|
397 |
+
elif output_bb_format == 'xyxy':
|
398 |
+
bb = [get_xyxy(item) for item in coord]
|
399 |
+
# print('bounding box!!!', bb)
|
400 |
+
return (text, bb), is_goal_filtered
|
401 |
+
|
402 |
+
|
403 |
+
|
weights/convert_safetensor_to_pt.py
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from ultralytics.nn.tasks import DetectionModel
|
3 |
+
from safetensors.torch import load_file
|
4 |
+
|
5 |
+
tensor_dict = load_file("weights/icon_detect/model.safetensors")
|
6 |
+
|
7 |
+
model = DetectionModel('weights/icon_detect/model.yaml')
|
8 |
+
model.load_state_dict(tensor_dict)
|
9 |
+
torch.save({'model':model}, 'weights/icon_detect/best.pt')
|
weights/icon_caption_florence/config.json
ADDED
@@ -0,0 +1,237 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "microsoft/Florence-2-base-ft",
|
3 |
+
"architectures": [
|
4 |
+
"Florence2ForConditionalGeneration"
|
5 |
+
],
|
6 |
+
"auto_map": {
|
7 |
+
"AutoConfig": "microsoft/Florence-2-base-ft--configuration_florence2.Florence2Config",
|
8 |
+
"AutoModelForCausalLM": "microsoft/Florence-2-base-ft--modeling_florence2.Florence2ForConditionalGeneration"
|
9 |
+
},
|
10 |
+
"bos_token_id": 2,
|
11 |
+
"eos_token_id": 1,
|
12 |
+
"ignore_index": -100,
|
13 |
+
"is_encoder_decoder": true,
|
14 |
+
"model_type": "florence2",
|
15 |
+
"pad_token_id": 0,
|
16 |
+
"projection_dim": 768,
|
17 |
+
"text_config": {
|
18 |
+
"_name_or_path": "",
|
19 |
+
"activation_dropout": 0.1,
|
20 |
+
"activation_function": "gelu",
|
21 |
+
"add_bias_logits": false,
|
22 |
+
"add_cross_attention": false,
|
23 |
+
"add_final_layer_norm": false,
|
24 |
+
"architectures": null,
|
25 |
+
"attention_dropout": 0.1,
|
26 |
+
"bad_words_ids": null,
|
27 |
+
"begin_suppress_tokens": null,
|
28 |
+
"bos_token_id": 0,
|
29 |
+
"chunk_size_feed_forward": 0,
|
30 |
+
"classif_dropout": 0.1,
|
31 |
+
"classifier_dropout": 0.0,
|
32 |
+
"cross_attention_hidden_size": null,
|
33 |
+
"d_model": 768,
|
34 |
+
"decoder_attention_heads": 12,
|
35 |
+
"decoder_ffn_dim": 3072,
|
36 |
+
"decoder_layerdrop": 0.0,
|
37 |
+
"decoder_layers": 6,
|
38 |
+
"decoder_start_token_id": 2,
|
39 |
+
"diversity_penalty": 0.0,
|
40 |
+
"do_sample": false,
|
41 |
+
"dropout": 0.1,
|
42 |
+
"early_stopping": true,
|
43 |
+
"encoder_attention_heads": 12,
|
44 |
+
"encoder_ffn_dim": 3072,
|
45 |
+
"encoder_layerdrop": 0.0,
|
46 |
+
"encoder_layers": 6,
|
47 |
+
"encoder_no_repeat_ngram_size": 0,
|
48 |
+
"eos_token_id": 2,
|
49 |
+
"exponential_decay_length_penalty": null,
|
50 |
+
"finetuning_task": null,
|
51 |
+
"forced_bos_token_id": 0,
|
52 |
+
"forced_eos_token_id": 2,
|
53 |
+
"gradient_checkpointing": false,
|
54 |
+
"id2label": {
|
55 |
+
"0": "LABEL_0",
|
56 |
+
"1": "LABEL_1",
|
57 |
+
"2": "LABEL_2"
|
58 |
+
},
|
59 |
+
"init_std": 0.02,
|
60 |
+
"is_decoder": false,
|
61 |
+
"is_encoder_decoder": true,
|
62 |
+
"label2id": {
|
63 |
+
"LABEL_0": 0,
|
64 |
+
"LABEL_1": 1,
|
65 |
+
"LABEL_2": 2
|
66 |
+
},
|
67 |
+
"length_penalty": 1.0,
|
68 |
+
"max_length": 20,
|
69 |
+
"max_position_embeddings": 1024,
|
70 |
+
"min_length": 0,
|
71 |
+
"model_type": "florence2_language",
|
72 |
+
"no_repeat_ngram_size": 3,
|
73 |
+
"normalize_before": false,
|
74 |
+
"num_beam_groups": 1,
|
75 |
+
"num_beams": 3,
|
76 |
+
"num_hidden_layers": 6,
|
77 |
+
"num_return_sequences": 1,
|
78 |
+
"output_attentions": false,
|
79 |
+
"output_hidden_states": false,
|
80 |
+
"output_scores": false,
|
81 |
+
"pad_token_id": 1,
|
82 |
+
"prefix": null,
|
83 |
+
"problem_type": null,
|
84 |
+
"pruned_heads": {},
|
85 |
+
"remove_invalid_values": false,
|
86 |
+
"repetition_penalty": 1.0,
|
87 |
+
"return_dict": true,
|
88 |
+
"return_dict_in_generate": false,
|
89 |
+
"scale_embedding": false,
|
90 |
+
"sep_token_id": null,
|
91 |
+
"suppress_tokens": null,
|
92 |
+
"task_specific_params": null,
|
93 |
+
"temperature": 1.0,
|
94 |
+
"tf_legacy_loss": false,
|
95 |
+
"tie_encoder_decoder": false,
|
96 |
+
"tie_word_embeddings": true,
|
97 |
+
"tokenizer_class": null,
|
98 |
+
"top_k": 50,
|
99 |
+
"top_p": 1.0,
|
100 |
+
"torch_dtype": null,
|
101 |
+
"torchscript": false,
|
102 |
+
"typical_p": 1.0,
|
103 |
+
"use_bfloat16": false,
|
104 |
+
"use_cache": true,
|
105 |
+
"vocab_size": 51289
|
106 |
+
},
|
107 |
+
"torch_dtype": "float32",
|
108 |
+
"transformers_version": "4.40.2",
|
109 |
+
"vision_config": {
|
110 |
+
"_name_or_path": "",
|
111 |
+
"add_cross_attention": false,
|
112 |
+
"architectures": null,
|
113 |
+
"bad_words_ids": null,
|
114 |
+
"begin_suppress_tokens": null,
|
115 |
+
"bos_token_id": null,
|
116 |
+
"chunk_size_feed_forward": 0,
|
117 |
+
"cross_attention_hidden_size": null,
|
118 |
+
"decoder_start_token_id": null,
|
119 |
+
"depths": [
|
120 |
+
1,
|
121 |
+
1,
|
122 |
+
9,
|
123 |
+
1
|
124 |
+
],
|
125 |
+
"dim_embed": [
|
126 |
+
128,
|
127 |
+
256,
|
128 |
+
512,
|
129 |
+
1024
|
130 |
+
],
|
131 |
+
"diversity_penalty": 0.0,
|
132 |
+
"do_sample": false,
|
133 |
+
"drop_path_rate": 0.1,
|
134 |
+
"early_stopping": false,
|
135 |
+
"enable_checkpoint": false,
|
136 |
+
"encoder_no_repeat_ngram_size": 0,
|
137 |
+
"eos_token_id": null,
|
138 |
+
"exponential_decay_length_penalty": null,
|
139 |
+
"finetuning_task": null,
|
140 |
+
"forced_bos_token_id": null,
|
141 |
+
"forced_eos_token_id": null,
|
142 |
+
"id2label": {
|
143 |
+
"0": "LABEL_0",
|
144 |
+
"1": "LABEL_1"
|
145 |
+
},
|
146 |
+
"image_feature_source": [
|
147 |
+
"spatial_avg_pool",
|
148 |
+
"temporal_avg_pool"
|
149 |
+
],
|
150 |
+
"image_pos_embed": {
|
151 |
+
"max_pos_embeddings": 50,
|
152 |
+
"type": "learned_abs_2d"
|
153 |
+
},
|
154 |
+
"is_decoder": false,
|
155 |
+
"is_encoder_decoder": false,
|
156 |
+
"label2id": {
|
157 |
+
"LABEL_0": 0,
|
158 |
+
"LABEL_1": 1
|
159 |
+
},
|
160 |
+
"length_penalty": 1.0,
|
161 |
+
"max_length": 20,
|
162 |
+
"min_length": 0,
|
163 |
+
"model_type": "davit",
|
164 |
+
"no_repeat_ngram_size": 0,
|
165 |
+
"num_beam_groups": 1,
|
166 |
+
"num_beams": 1,
|
167 |
+
"num_groups": [
|
168 |
+
4,
|
169 |
+
8,
|
170 |
+
16,
|
171 |
+
32
|
172 |
+
],
|
173 |
+
"num_heads": [
|
174 |
+
4,
|
175 |
+
8,
|
176 |
+
16,
|
177 |
+
32
|
178 |
+
],
|
179 |
+
"num_return_sequences": 1,
|
180 |
+
"output_attentions": false,
|
181 |
+
"output_hidden_states": false,
|
182 |
+
"output_scores": false,
|
183 |
+
"pad_token_id": null,
|
184 |
+
"patch_padding": [
|
185 |
+
3,
|
186 |
+
1,
|
187 |
+
1,
|
188 |
+
1
|
189 |
+
],
|
190 |
+
"patch_prenorm": [
|
191 |
+
false,
|
192 |
+
true,
|
193 |
+
true,
|
194 |
+
true
|
195 |
+
],
|
196 |
+
"patch_size": [
|
197 |
+
7,
|
198 |
+
3,
|
199 |
+
3,
|
200 |
+
3
|
201 |
+
],
|
202 |
+
"patch_stride": [
|
203 |
+
4,
|
204 |
+
2,
|
205 |
+
2,
|
206 |
+
2
|
207 |
+
],
|
208 |
+
"prefix": null,
|
209 |
+
"problem_type": null,
|
210 |
+
"projection_dim": 768,
|
211 |
+
"pruned_heads": {},
|
212 |
+
"remove_invalid_values": false,
|
213 |
+
"repetition_penalty": 1.0,
|
214 |
+
"return_dict": true,
|
215 |
+
"return_dict_in_generate": false,
|
216 |
+
"sep_token_id": null,
|
217 |
+
"suppress_tokens": null,
|
218 |
+
"task_specific_params": null,
|
219 |
+
"temperature": 1.0,
|
220 |
+
"tf_legacy_loss": false,
|
221 |
+
"tie_encoder_decoder": false,
|
222 |
+
"tie_word_embeddings": true,
|
223 |
+
"tokenizer_class": null,
|
224 |
+
"top_k": 50,
|
225 |
+
"top_p": 1.0,
|
226 |
+
"torch_dtype": null,
|
227 |
+
"torchscript": false,
|
228 |
+
"typical_p": 1.0,
|
229 |
+
"use_bfloat16": false,
|
230 |
+
"visual_temporal_embedding": {
|
231 |
+
"max_temporal_embeddings": 100,
|
232 |
+
"type": "COSINE"
|
233 |
+
},
|
234 |
+
"window_size": 12
|
235 |
+
},
|
236 |
+
"vocab_size": 51289
|
237 |
+
}
|
weights/icon_caption_florence/generation_config.json
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_from_model_config": true,
|
3 |
+
"bos_token_id": 2,
|
4 |
+
"decoder_start_token_id": 2,
|
5 |
+
"early_stopping": true,
|
6 |
+
"eos_token_id": 1,
|
7 |
+
"forced_bos_token_id": 0,
|
8 |
+
"forced_eos_token_id": 2,
|
9 |
+
"no_repeat_ngram_size": 3,
|
10 |
+
"num_beams": 3,
|
11 |
+
"pad_token_id": 0,
|
12 |
+
"transformers_version": "4.40.2"
|
13 |
+
}
|
weights/icon_caption_florence/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c0d7a6bedf8c6dbabe7d40b9f78ada36e78c1a93617506bc06a93279a78dfb14
|
3 |
+
size 1083916964
|
weights/icon_detect/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:84caa6c7e0607d7d6edddb46f2affd0d0d86ddf34e71fbcce7b4ba461bd97574
|
3 |
+
size 6075790
|
weights/icon_detect/model.yaml
ADDED
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
backbone:
|
2 |
+
- - -1
|
3 |
+
- 1
|
4 |
+
- Conv
|
5 |
+
- - 64
|
6 |
+
- 3
|
7 |
+
- 2
|
8 |
+
- - -1
|
9 |
+
- 1
|
10 |
+
- Conv
|
11 |
+
- - 128
|
12 |
+
- 3
|
13 |
+
- 2
|
14 |
+
- - -1
|
15 |
+
- 3
|
16 |
+
- C2f
|
17 |
+
- - 128
|
18 |
+
- true
|
19 |
+
- - -1
|
20 |
+
- 1
|
21 |
+
- Conv
|
22 |
+
- - 256
|
23 |
+
- 3
|
24 |
+
- 2
|
25 |
+
- - -1
|
26 |
+
- 6
|
27 |
+
- C2f
|
28 |
+
- - 256
|
29 |
+
- true
|
30 |
+
- - -1
|
31 |
+
- 1
|
32 |
+
- Conv
|
33 |
+
- - 512
|
34 |
+
- 3
|
35 |
+
- 2
|
36 |
+
- - -1
|
37 |
+
- 6
|
38 |
+
- C2f
|
39 |
+
- - 512
|
40 |
+
- true
|
41 |
+
- - -1
|
42 |
+
- 1
|
43 |
+
- Conv
|
44 |
+
- - 1024
|
45 |
+
- 3
|
46 |
+
- 2
|
47 |
+
- - -1
|
48 |
+
- 3
|
49 |
+
- C2f
|
50 |
+
- - 1024
|
51 |
+
- true
|
52 |
+
- - -1
|
53 |
+
- 1
|
54 |
+
- SPPF
|
55 |
+
- - 1024
|
56 |
+
- 5
|
57 |
+
ch: 3
|
58 |
+
depth_multiple: 0.33
|
59 |
+
head:
|
60 |
+
- - -1
|
61 |
+
- 1
|
62 |
+
- nn.Upsample
|
63 |
+
- - None
|
64 |
+
- 2
|
65 |
+
- nearest
|
66 |
+
- - - -1
|
67 |
+
- 6
|
68 |
+
- 1
|
69 |
+
- Concat
|
70 |
+
- - 1
|
71 |
+
- - -1
|
72 |
+
- 3
|
73 |
+
- C2f
|
74 |
+
- - 512
|
75 |
+
- - -1
|
76 |
+
- 1
|
77 |
+
- nn.Upsample
|
78 |
+
- - None
|
79 |
+
- 2
|
80 |
+
- nearest
|
81 |
+
- - - -1
|
82 |
+
- 4
|
83 |
+
- 1
|
84 |
+
- Concat
|
85 |
+
- - 1
|
86 |
+
- - -1
|
87 |
+
- 3
|
88 |
+
- C2f
|
89 |
+
- - 256
|
90 |
+
- - -1
|
91 |
+
- 1
|
92 |
+
- Conv
|
93 |
+
- - 256
|
94 |
+
- 3
|
95 |
+
- 2
|
96 |
+
- - - -1
|
97 |
+
- 12
|
98 |
+
- 1
|
99 |
+
- Concat
|
100 |
+
- - 1
|
101 |
+
- - -1
|
102 |
+
- 3
|
103 |
+
- C2f
|
104 |
+
- - 512
|
105 |
+
- - -1
|
106 |
+
- 1
|
107 |
+
- Conv
|
108 |
+
- - 512
|
109 |
+
- 3
|
110 |
+
- 2
|
111 |
+
- - - -1
|
112 |
+
- 9
|
113 |
+
- 1
|
114 |
+
- Concat
|
115 |
+
- - 1
|
116 |
+
- - -1
|
117 |
+
- 3
|
118 |
+
- C2f
|
119 |
+
- - 1024
|
120 |
+
- - - 15
|
121 |
+
- 18
|
122 |
+
- 21
|
123 |
+
- 1
|
124 |
+
- Detect
|
125 |
+
- - nc
|
126 |
+
nc: 1
|
127 |
+
width_multiple: 0.25
|