Update Action-Effect.py
Browse files- Action-Effect.py +32 -3
Action-Effect.py
CHANGED
@@ -1,6 +1,10 @@
|
|
1 |
import json
|
2 |
from zipfile import ZipFile
|
3 |
import datasets
|
|
|
|
|
|
|
|
|
4 |
|
5 |
_CITATION = """\
|
6 |
@inproceedings{gao-etal-2018-action,
|
@@ -26,6 +30,9 @@ Despite recent advances in knowledge representation, automated reasoning, and ma
|
|
26 |
|
27 |
_URL = "https://huggingface.co/datasets/sled-umich/Action-Effect"
|
28 |
|
|
|
|
|
|
|
29 |
|
30 |
class Action_Effect(datasets.GeneratorBasedBuilder):
|
31 |
"""
|
@@ -42,6 +49,8 @@ class Action_Effect(datasets.GeneratorBasedBuilder):
|
|
42 |
"verb noun": datasets.Value("string"),
|
43 |
"effect_sentence_list": datasets.features.Sequence(datasets.Value("string")),
|
44 |
"effect_phrase_list": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("string"))),
|
|
|
|
|
45 |
}
|
46 |
),
|
47 |
homepage=_URL,
|
@@ -49,18 +58,38 @@ class Action_Effect(datasets.GeneratorBasedBuilder):
|
|
49 |
)
|
50 |
|
51 |
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
|
|
|
|
|
|
52 |
return [
|
53 |
datasets.SplitGenerator(
|
54 |
-
name="ActionEffect",
|
55 |
gen_kwargs={
|
56 |
-
"action_effect_info_path":
|
57 |
},
|
58 |
)
|
59 |
]
|
60 |
|
61 |
-
def _generate_examples(self, action_effect_info_path):
|
62 |
with open(action_effect_info_path) as f:
|
63 |
action_effect_info = json.load(f)
|
|
|
|
|
|
|
|
|
64 |
for idx, this_ae_info in enumerate(action_effect_info):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
yield idx, this_ae_info
|
66 |
|
|
|
1 |
import json
|
2 |
from zipfile import ZipFile
|
3 |
import datasets
|
4 |
+
import os
|
5 |
+
from PIL import Image
|
6 |
+
from numpy import asarray
|
7 |
+
import tarfile
|
8 |
|
9 |
_CITATION = """\
|
10 |
@inproceedings{gao-etal-2018-action,
|
|
|
30 |
|
31 |
_URL = "https://huggingface.co/datasets/sled-umich/Action-Effect"
|
32 |
|
33 |
+
JSON_URL = "http://162.212.153.129/action_effect_info.json"
|
34 |
+
|
35 |
+
IMGS_URL = "action_effect_image_rs.tar.gz"
|
36 |
|
37 |
class Action_Effect(datasets.GeneratorBasedBuilder):
|
38 |
"""
|
|
|
49 |
"verb noun": datasets.Value("string"),
|
50 |
"effect_sentence_list": datasets.features.Sequence(datasets.Value("string")),
|
51 |
"effect_phrase_list": datasets.features.Sequence(datasets.features.Sequence(datasets.Value("string"))),
|
52 |
+
"positvie_image_list": datasets.features.Sequence(datasets.Image()),
|
53 |
+
"negative_image_list": datasets.features.Sequence(datasets.Image()),
|
54 |
}
|
55 |
),
|
56 |
homepage=_URL,
|
|
|
58 |
)
|
59 |
|
60 |
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
61 |
+
dl_dir = dl_manager.download_and_extract(JSON_URL)
|
62 |
+
action_effect_info_path = dl_dir + "/action_effect_info.json"
|
63 |
+
img_zip_path = dl_dir + "/action_effect_image_rs.tar.gz"
|
64 |
return [
|
65 |
datasets.SplitGenerator(
|
66 |
+
name= "ActionEffect",
|
67 |
gen_kwargs={
|
68 |
+
"action_effect_info_path": action_effect_info_path, "img_zip_path": img_zip_path
|
69 |
},
|
70 |
)
|
71 |
]
|
72 |
|
73 |
+
def _generate_examples(self, action_effect_info_path, img_zip_path):
|
74 |
with open(action_effect_info_path) as f:
|
75 |
action_effect_info = json.load(f)
|
76 |
+
img_zip = tarfile.open(img_zip_path)
|
77 |
+
img_zip.extractall("./action_effect_image_rs")
|
78 |
+
img_root = "./action_effect_image_rs/"
|
79 |
+
img_zip.close()
|
80 |
for idx, this_ae_info in enumerate(action_effect_info):
|
81 |
+
this_ae_info['positive_image_list'] = []
|
82 |
+
this_ae_info['negative_image_list'] = []
|
83 |
+
vn = this_ae_info["verb noun"]
|
84 |
+
this_image_root_positive = img_root + vn.replace(" ", "+") + "/positive/"
|
85 |
+
this_image_root_negative = img_root + vn.replace(" ", "+") + "/positive/"
|
86 |
+
for img_name in os.listdir(this_image_root_positive):
|
87 |
+
img_pil = Image.open(this_image_root_positive + img_name)
|
88 |
+
img_np = asarray(img_pil)
|
89 |
+
this_image_root_positive.append(img_np)
|
90 |
+
for img_name in os.listdir(this_image_root_negative):
|
91 |
+
img_pil = Image.open(this_image_root_negative + img_name)
|
92 |
+
img_np = asarray(img_pil)
|
93 |
+
this_image_root_negative.append(img_np)
|
94 |
yield idx, this_ae_info
|
95 |
|