Hugging Face's logo

qualcomm
/
FFNet-78S

Image Segmentation
PyTorch
TF Lite
ONNX
android
Model card Files Community
qaihm-bot commited on
Commit
2c76d38
·
verified ·
1 Parent(s): 1019644

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +40 -19
README.md CHANGED
@@ -16,7 +16,7 @@ tags:
16
 
17
  FFNet-78S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.
18
 
19
- This model is an implementation of FFNet-78S found [here](https://github.com/Qualcomm-AI-research/FFNet).
20
  This repository provides scripts to run FFNet-78S on Qualcomm® devices.
21
  More details on model performance across various devices, can be found
22
  [here](https://aihub.qualcomm.com/models/ffnet_78s).
@@ -32,15 +32,32 @@ More details on model performance across various devices, can be found
32
  - Model size: 105 MB
33
  - Number of output classes: 19
34
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
 
36
 
37
 
38
- | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
39
- | ---|---|---|---|---|---|---|---|
40
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | TFLite | 23.714 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite)
41
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 23.669 ms | 2 - 24 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so)
42
-
43
-
44
 
45
  ## Installation
46
 
@@ -96,16 +113,16 @@ device. This script does the following:
96
  ```bash
97
  python -m qai_hub_models.models.ffnet_78s.export
98
  ```
99
-
100
  ```
101
- Profile Job summary of FFNet-78S
102
- --------------------------------------------------
103
- Device: Snapdragon X Elite CRD (11)
104
- Estimated Inference Time: 24.05 ms
105
- Estimated Peak Memory Range: 24.05-24.05 MB
106
- Compute Units: NPU (235) | Total (235)
107
-
108
-
 
109
  ```
110
 
111
 
@@ -204,15 +221,19 @@ provides instructions on how to use the `.so` shared library in an Android appl
204
  Get more details on FFNet-78S's performance across various devices [here](https://aihub.qualcomm.com/models/ffnet_78s).
205
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
206
 
 
207
  ## License
208
- - The license for the original implementation of FFNet-78S can be found
209
- [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
210
- - The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
 
211
 
212
  ## References
213
  * [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
214
  * [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)
215
 
 
 
216
  ## Community
217
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
218
  * For questions or feedback please [reach out to us](mailto:[email protected]).
 
16
 
17
  FFNet-78S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.
18
 
19
+ This model is an implementation of FFNet-78S found [here]({source_repo}).
20
  This repository provides scripts to run FFNet-78S on Qualcomm® devices.
21
  More details on model performance across various devices, can be found
22
  [here](https://aihub.qualcomm.com/models/ffnet_78s).
 
32
  - Model size: 105 MB
33
  - Number of output classes: 19
34
 
35
+ | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
36
+ |---|---|---|---|---|---|---|---|---|
37
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 23.254 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
38
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 23.635 ms | 24 - 48 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
39
+ | FFNet-78S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 32.7 ms | 24 - 55 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
40
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 21.162 ms | 2 - 130 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
41
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 21.44 ms | 20 - 61 MB | FP16 | NPU | [FFNet-78S.so](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.so) |
42
+ | FFNet-78S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 29.103 ms | 2 - 153 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
43
+ | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 23.104 ms | 2 - 5 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
44
+ | FFNet-78S | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 23.037 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
45
+ | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 23.077 ms | 2 - 5 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
46
+ | FFNet-78S | SA8255 (Proxy) | SA8255P Proxy | QNN | 23.073 ms | 24 - 26 MB | FP16 | NPU | Use Export Script |
47
+ | FFNet-78S | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 23.169 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
48
+ | FFNet-78S | SA8775 (Proxy) | SA8775P Proxy | QNN | 23.407 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
49
+ | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 23.24 ms | 2 - 4 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
50
+ | FFNet-78S | SA8650 (Proxy) | SA8650P Proxy | QNN | 23.501 ms | 24 - 28 MB | FP16 | NPU | Use Export Script |
51
+ | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 39.261 ms | 1 - 108 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
52
+ | FFNet-78S | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 39.234 ms | 24 - 58 MB | FP16 | NPU | Use Export Script |
53
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 16.614 ms | 2 - 55 MB | FP16 | NPU | [FFNet-78S.tflite](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.tflite) |
54
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 16.705 ms | 24 - 64 MB | FP16 | NPU | Use Export Script |
55
+ | FFNet-78S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 20.857 ms | 26 - 86 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
56
+ | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 23.035 ms | 24 - 24 MB | FP16 | NPU | Use Export Script |
57
+ | FFNet-78S | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 36.44 ms | 31 - 31 MB | FP16 | NPU | [FFNet-78S.onnx](https://huggingface.co/qualcomm/FFNet-78S/blob/main/FFNet-78S.onnx) |
58
 
59
 
60
 
 
 
 
 
 
 
61
 
62
  ## Installation
63
 
 
113
  ```bash
114
  python -m qai_hub_models.models.ffnet_78s.export
115
  ```
 
116
  ```
117
+ Profiling Results
118
+ ------------------------------------------------------------
119
+ FFNet-78S
120
+ Device : Samsung Galaxy S23 (13)
121
+ Runtime : TFLITE
122
+ Estimated inference time (ms) : 23.3
123
+ Estimated peak memory usage (MB): [2, 4]
124
+ Total # Ops : 149
125
+ Compute Unit(s) : NPU (149 ops)
126
  ```
127
 
128
 
 
221
  Get more details on FFNet-78S's performance across various devices [here](https://aihub.qualcomm.com/models/ffnet_78s).
222
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
223
 
224
+
225
  ## License
226
+ * The license for the original implementation of FFNet-78S can be found [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
227
+ * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
228
+
229
+
230
 
231
  ## References
232
  * [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
233
  * [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)
234
 
235
+
236
+
237
  ## Community
238
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
239
  * For questions or feedback please [reach out to us](mailto:[email protected]).