Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
@@ -14,7 +14,7 @@ tags:
|
|
14 |
|
15 |
The Fast Segment Anything Model (FastSAM) is a novel, real-time CNN-based solution for the Segment Anything task. This task is designed to segment any object within an image based on various possible user interaction prompts. The model performs competitively despite significantly reduced computation, making it a practical choice for a variety of vision tasks.
|
16 |
|
17 |
-
This model is an implementation of FastSam-S found [here](
|
18 |
This repository provides scripts to run FastSam-S on Qualcomm® devices.
|
19 |
More details on model performance across various devices, can be found
|
20 |
[here](https://aihub.qualcomm.com/models/fastsam_s).
|
@@ -30,14 +30,24 @@ More details on model performance across various devices, can be found
|
|
30 |
- Number of parameters: 11.8M
|
31 |
- Model size: 45.1 MB
|
32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
|
35 |
|
36 |
-
| Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
37 |
-
| ---|---|---|---|---|---|---|---|
|
38 |
-
| Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 8.062 ms | 6 - 22 MB | FP16 | NPU | [FastSam-S.so](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.so)
|
39 |
-
|
40 |
-
|
41 |
|
42 |
## Installation
|
43 |
|
@@ -93,16 +103,16 @@ device. This script does the following:
|
|
93 |
```bash
|
94 |
python -m qai_hub_models.models.fastsam_s.export
|
95 |
```
|
96 |
-
|
97 |
```
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
|
|
106 |
```
|
107 |
|
108 |
|
@@ -201,15 +211,19 @@ provides instructions on how to use the `.so` shared library in an Android appl
|
|
201 |
Get more details on FastSam-S's performance across various devices [here](https://aihub.qualcomm.com/models/fastsam_s).
|
202 |
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
|
203 |
|
|
|
204 |
## License
|
205 |
-
|
206 |
-
|
207 |
-
|
|
|
208 |
|
209 |
## References
|
210 |
* [Fast Segment Anything](https://arxiv.org/abs/2306.12156)
|
211 |
* [Source Model Implementation](https://github.com/CASIA-IVA-Lab/FastSAM)
|
212 |
|
|
|
|
|
213 |
## Community
|
214 |
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
|
215 |
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
|
|
|
14 |
|
15 |
The Fast Segment Anything Model (FastSAM) is a novel, real-time CNN-based solution for the Segment Anything task. This task is designed to segment any object within an image based on various possible user interaction prompts. The model performs competitively despite significantly reduced computation, making it a practical choice for a variety of vision tasks.
|
16 |
|
17 |
+
This model is an implementation of FastSam-S found [here]({source_repo}).
|
18 |
This repository provides scripts to run FastSam-S on Qualcomm® devices.
|
19 |
More details on model performance across various devices, can be found
|
20 |
[here](https://aihub.qualcomm.com/models/fastsam_s).
|
|
|
30 |
- Number of parameters: 11.8M
|
31 |
- Model size: 45.1 MB
|
32 |
|
33 |
+
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
|
34 |
+
|---|---|---|---|---|---|---|---|---|
|
35 |
+
| FastSam-S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 8.064 ms | 4 - 18 MB | FP16 | NPU | [FastSam-S.so](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.so) |
|
36 |
+
| FastSam-S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 9.58 ms | 4 - 25 MB | FP16 | NPU | [FastSam-S.onnx](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.onnx) |
|
37 |
+
| FastSam-S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 6.96 ms | 5 - 38 MB | FP16 | NPU | [FastSam-S.so](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.so) |
|
38 |
+
| FastSam-S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 7.273 ms | 1 - 81 MB | FP16 | NPU | [FastSam-S.onnx](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.onnx) |
|
39 |
+
| FastSam-S | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 7.38 ms | 5 - 10 MB | FP16 | NPU | Use Export Script |
|
40 |
+
| FastSam-S | SA8255 (Proxy) | SA8255P Proxy | QNN | 7.689 ms | 5 - 10 MB | FP16 | NPU | Use Export Script |
|
41 |
+
| FastSam-S | SA8775 (Proxy) | SA8775P Proxy | QNN | 7.719 ms | 5 - 9 MB | FP16 | NPU | Use Export Script |
|
42 |
+
| FastSam-S | SA8650 (Proxy) | SA8650P Proxy | QNN | 7.618 ms | 5 - 8 MB | FP16 | NPU | Use Export Script |
|
43 |
+
| FastSam-S | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 13.749 ms | 5 - 42 MB | FP16 | NPU | Use Export Script |
|
44 |
+
| FastSam-S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 5.49 ms | 5 - 36 MB | FP16 | NPU | Use Export Script |
|
45 |
+
| FastSam-S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 5.354 ms | 16 - 62 MB | FP16 | NPU | [FastSam-S.onnx](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.onnx) |
|
46 |
+
| FastSam-S | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 8.317 ms | 5 - 5 MB | FP16 | NPU | Use Export Script |
|
47 |
+
| FastSam-S | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 9.903 ms | 21 - 21 MB | FP16 | NPU | [FastSam-S.onnx](https://huggingface.co/qualcomm/FastSam-S/blob/main/FastSam-S.onnx) |
|
48 |
|
49 |
|
50 |
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
## Installation
|
53 |
|
|
|
103 |
```bash
|
104 |
python -m qai_hub_models.models.fastsam_s.export
|
105 |
```
|
|
|
106 |
```
|
107 |
+
Profiling Results
|
108 |
+
------------------------------------------------------------
|
109 |
+
FastSam-S
|
110 |
+
Device : Samsung Galaxy S23 (13)
|
111 |
+
Runtime : QNN
|
112 |
+
Estimated inference time (ms) : 8.1
|
113 |
+
Estimated peak memory usage (MB): [4, 18]
|
114 |
+
Total # Ops : 286
|
115 |
+
Compute Unit(s) : NPU (286 ops)
|
116 |
```
|
117 |
|
118 |
|
|
|
211 |
Get more details on FastSam-S's performance across various devices [here](https://aihub.qualcomm.com/models/fastsam_s).
|
212 |
Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
|
213 |
|
214 |
+
|
215 |
## License
|
216 |
+
* The license for the original implementation of FastSam-S can be found [here](https://github.com/CASIA-IVA-Lab/FastSAM/blob/main/LICENSE).
|
217 |
+
* The license for the compiled assets for on-device deployment can be found [here](https://github.com/CASIA-IVA-Lab/FastSAM/blob/main/LICENSE)
|
218 |
+
|
219 |
+
|
220 |
|
221 |
## References
|
222 |
* [Fast Segment Anything](https://arxiv.org/abs/2306.12156)
|
223 |
* [Source Model Implementation](https://github.com/CASIA-IVA-Lab/FastSAM)
|
224 |
|
225 |
+
|
226 |
+
|
227 |
## Community
|
228 |
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
|
229 |
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
|