Hugging Face's logo

qualcomm
/
FFNet-40S

Image Segmentation
PyTorch
LiteRT
real_time
android
Model card Files
xet
Community
qaihm-bot commited on
Commit
1d60f3e
·
verified ·
1 Parent(s): f209fc3

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +40 -19
README.md CHANGED
@@ -17,7 +17,7 @@ tags:
17
 
18
  FFNet-40S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.
19
 
20
- This model is an implementation of FFNet-40S found [here](https://github.com/Qualcomm-AI-research/FFNet).
21
  This repository provides scripts to run FFNet-40S on Qualcomm® devices.
22
  More details on model performance across various devices, can be found
23
  [here](https://aihub.qualcomm.com/models/ffnet_40s).
@@ -33,15 +33,32 @@ More details on model performance across various devices, can be found
33
  - Model size: 53.1 MB
34
  - Number of output classes: 19
35
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
36
 
37
 
38
 
39
- | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
40
- | ---|---|---|---|---|---|---|---|
41
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | TFLite | 17.077 ms | 2 - 4 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite)
42
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 17.566 ms | 26 - 40 MB | FP16 | NPU | [FFNet-40S.so](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.so)
43
-
44
-
45
 
46
  ## Installation
47
 
@@ -97,16 +114,16 @@ device. This script does the following:
97
  ```bash
98
  python -m qai_hub_models.models.ffnet_40s.export
99
  ```
100
-
101
  ```
102
- Profile Job summary of FFNet-40S
103
- --------------------------------------------------
104
- Device: Snapdragon X Elite CRD (11)
105
- Estimated Inference Time: 17.86 ms
106
- Estimated Peak Memory Range: 24.05-24.05 MB
107
- Compute Units: NPU (140) | Total (140)
108
-
109
-
 
110
  ```
111
 
112
 
@@ -205,15 +222,19 @@ provides instructions on how to use the `.so` shared library in an Android appl
205
  Get more details on FFNet-40S's performance across various devices [here](https://aihub.qualcomm.com/models/ffnet_40s).
206
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
207
 
 
208
  ## License
209
- - The license for the original implementation of FFNet-40S can be found
210
- [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
211
- - 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)
 
212
 
213
  ## References
214
  * [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
215
  * [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)
216
 
 
 
217
  ## Community
218
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
219
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).
 
17
 
18
  FFNet-40S is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.
19
 
20
+ This model is an implementation of FFNet-40S found [here]({source_repo}).
21
  This repository provides scripts to run FFNet-40S on Qualcomm® devices.
22
  More details on model performance across various devices, can be found
23
  [here](https://aihub.qualcomm.com/models/ffnet_40s).
 
33
  - Model size: 53.1 MB
34
  - Number of output classes: 19
35
 
36
+ | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
+ |---|---|---|---|---|---|---|---|---|
38
+ | FFNet-40S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 17.007 ms | 2 - 5 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
39
+ | FFNet-40S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 17.621 ms | 25 - 49 MB | FP16 | NPU | [FFNet-40S.so](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.so) |
40
+ | FFNet-40S | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 24.964 ms | 26 - 28 MB | FP16 | NPU | [FFNet-40S.onnx](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.onnx) |
41
+ | FFNet-40S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 14.889 ms | 2 - 102 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
42
+ | FFNet-40S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 15.114 ms | 24 - 57 MB | FP16 | NPU | [FFNet-40S.so](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.so) |
43
+ | FFNet-40S | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 22.02 ms | 27 - 151 MB | FP16 | NPU | [FFNet-40S.onnx](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.onnx) |
44
+ | FFNet-40S | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 16.785 ms | 2 - 5 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
45
+ | FFNet-40S | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 16.327 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
46
+ | FFNet-40S | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 16.799 ms | 2 - 4 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
47
+ | FFNet-40S | SA8255 (Proxy) | SA8255P Proxy | QNN | 16.511 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
48
+ | FFNet-40S | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 16.774 ms | 2 - 5 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
49
+ | FFNet-40S | SA8775 (Proxy) | SA8775P Proxy | QNN | 16.85 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
50
+ | FFNet-40S | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 16.854 ms | 2 - 5 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
51
+ | FFNet-40S | SA8650 (Proxy) | SA8650P Proxy | QNN | 16.816 ms | 24 - 25 MB | FP16 | NPU | Use Export Script |
52
+ | FFNet-40S | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 27.915 ms | 2 - 93 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
53
+ | FFNet-40S | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 28.352 ms | 22 - 55 MB | FP16 | NPU | Use Export Script |
54
+ | FFNet-40S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 11.794 ms | 1 - 44 MB | FP16 | NPU | [FFNet-40S.tflite](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.tflite) |
55
+ | FFNet-40S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 12.082 ms | 24 - 56 MB | FP16 | NPU | Use Export Script |
56
+ | FFNet-40S | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 15.466 ms | 32 - 84 MB | FP16 | NPU | [FFNet-40S.onnx](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.onnx) |
57
+ | FFNet-40S | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 16.542 ms | 24 - 24 MB | FP16 | NPU | Use Export Script |
58
+ | FFNet-40S | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 30.278 ms | 24 - 24 MB | FP16 | NPU | [FFNet-40S.onnx](https://huggingface.co/qualcomm/FFNet-40S/blob/main/FFNet-40S.onnx) |
59
 
60
 
61
 
 
 
 
 
 
 
62
 
63
  ## Installation
64
 
 
114
  ```bash
115
  python -m qai_hub_models.models.ffnet_40s.export
116
  ```
 
117
  ```
118
+ Profiling Results
119
+ ------------------------------------------------------------
120
+ FFNet-40S
121
+ Device : Samsung Galaxy S23 (13)
122
+ Runtime : TFLITE
123
+ Estimated inference time (ms) : 17.0
124
+ Estimated peak memory usage (MB): [2, 5]
125
+ Total # Ops : 92
126
+ Compute Unit(s) : NPU (92 ops)
127
  ```
128
 
129
 
 
222
  Get more details on FFNet-40S's performance across various devices [here](https://aihub.qualcomm.com/models/ffnet_40s).
223
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
224
 
225
+
226
  ## License
227
+ * The license for the original implementation of FFNet-40S can be found [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
228
+ * 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)
229
+
230
+
231
 
232
  ## References
233
  * [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
234
  * [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)
235
 
236
+
237
+
238
  ## Community
239
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
240
  * For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).