Hugging Face's logo

qualcomm
/
DeepLabV3-Plus-MobileNet-Quantized

Image Segmentation
PyTorch
TF Lite
ONNX
quantized
android
Model card Files Community
qaihm-bot commited on
Commit
105a7a6
•
1 Parent(s): 073e8df

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +43 -19
README.md CHANGED
@@ -17,7 +17,7 @@ tags:
17
 
18
  DeepLabV3 Quantized is designed for semantic segmentation at multiple scales, trained on various datasets. It uses MobileNet as a backbone.
19
 
20
- This model is an implementation of DeepLabV3-Plus-MobileNet-Quantized found [here](https://github.com/jfzhang95/pytorch-deeplab-xception).
21
  This repository provides scripts to run DeepLabV3-Plus-MobileNet-Quantized on Qualcomm® devices.
22
  More details on model performance across various devices, can be found
23
  [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet_quantized).
@@ -33,15 +33,35 @@ More details on model performance across various devices, can be found
33
  - Model size: 6.04 MB
34
  - Number of output classes: 21
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 | 3.353 ms | 0 - 8 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite)
42
- | Samsung Galaxy S23 Ultra (Android 13) | Snapdragon® 8 Gen 2 | QNN Model Library | 5.163 ms | 0 - 14 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.so](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.so)
43
-
44
-
45
 
46
  ## Installation
47
 
@@ -97,16 +117,16 @@ device. This script does the following:
97
  ```bash
98
  python -m qai_hub_models.models.deeplabv3_plus_mobilenet_quantized.export
99
  ```
100
-
101
  ```
102
- Profile Job summary of DeepLabV3-Plus-MobileNet-Quantized
103
- --------------------------------------------------
104
- Device: Snapdragon X Elite CRD (11)
105
- Estimated Inference Time: 4.29 ms
106
- Estimated Peak Memory Range: 0.77-0.77 MB
107
- Compute Units: NPU (99) | Total (99)
108
-
109
-
 
110
  ```
111
 
112
 
@@ -145,15 +165,19 @@ provides instructions on how to use the `.so` shared library in an Android appl
145
  Get more details on DeepLabV3-Plus-MobileNet-Quantized's performance across various devices [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet_quantized).
146
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
147
 
 
148
  ## License
149
- - The license for the original implementation of DeepLabV3-Plus-MobileNet-Quantized can be found
150
- [here](https://github.com/quic/aimet-model-zoo/blob/develop/LICENSE.pdf).
151
- - 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)
 
152
 
153
  ## References
154
  * [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1706.05587)
155
  * [Source Model Implementation](https://github.com/jfzhang95/pytorch-deeplab-xception)
156
 
 
 
157
  ## Community
158
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
159
  * For questions or feedback please [reach out to us](mailto:[email protected]).
 
17
 
18
  DeepLabV3 Quantized is designed for semantic segmentation at multiple scales, trained on various datasets. It uses MobileNet as a backbone.
19
 
20
+ This model is an implementation of DeepLabV3-Plus-MobileNet-Quantized found [here]({source_repo}).
21
  This repository provides scripts to run DeepLabV3-Plus-MobileNet-Quantized on Qualcomm® devices.
22
  More details on model performance across various devices, can be found
23
  [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet_quantized).
 
33
  - Model size: 6.04 MB
34
  - Number of output classes: 21
35
 
36
+ | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
37
+ |---|---|---|---|---|---|---|---|---|
38
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 3.304 ms | 0 - 146 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
39
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 5.214 ms | 0 - 12 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.so](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.so) |
40
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 4.221 ms | 11 - 18 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.onnx) |
41
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 2.825 ms | 0 - 65 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
42
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 3.844 ms | 1 - 25 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.so](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.so) |
43
+ | DeepLabV3-Plus-MobileNet-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 3.141 ms | 0 - 72 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.onnx) |
44
+ | DeepLabV3-Plus-MobileNet-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 14.162 ms | 5 - 48 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
45
+ | DeepLabV3-Plus-MobileNet-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 18.291 ms | 1 - 9 MB | INT8 | NPU | Use Export Script |
46
+ | DeepLabV3-Plus-MobileNet-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 127.38 ms | 11 - 63 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
47
+ | DeepLabV3-Plus-MobileNet-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 3.315 ms | 0 - 8 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
48
+ | DeepLabV3-Plus-MobileNet-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 3.963 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
49
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 3.335 ms | 0 - 4 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
50
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 3.97 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
51
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8775 (Proxy) | SA8775P Proxy | TFLITE | 3.294 ms | 0 - 9 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
52
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8775 (Proxy) | SA8775P Proxy | QNN | 3.994 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
53
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 3.328 ms | 0 - 115 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
54
+ | DeepLabV3-Plus-MobileNet-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 3.963 ms | 1 - 2 MB | INT8 | NPU | Use Export Script |
55
+ | DeepLabV3-Plus-MobileNet-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 4.166 ms | 5 - 71 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
56
+ | DeepLabV3-Plus-MobileNet-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 5.51 ms | 1 - 32 MB | INT8 | NPU | Use Export Script |
57
+ | DeepLabV3-Plus-MobileNet-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 2.441 ms | 0 - 42 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.tflite](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.tflite) |
58
+ | DeepLabV3-Plus-MobileNet-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 3.816 ms | 1 - 25 MB | INT8 | NPU | Use Export Script |
59
+ | DeepLabV3-Plus-MobileNet-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 2.494 ms | 0 - 47 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.onnx) |
60
+ | DeepLabV3-Plus-MobileNet-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 4.324 ms | 1 - 1 MB | INT8 | NPU | Use Export Script |
61
+ | DeepLabV3-Plus-MobileNet-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 4.68 ms | 17 - 17 MB | INT8 | NPU | [DeepLabV3-Plus-MobileNet-Quantized.onnx](https://huggingface.co/qualcomm/DeepLabV3-Plus-MobileNet-Quantized/blob/main/DeepLabV3-Plus-MobileNet-Quantized.onnx) |
62
 
63
 
64
 
 
 
 
 
 
 
65
 
66
  ## Installation
67
 
 
117
  ```bash
118
  python -m qai_hub_models.models.deeplabv3_plus_mobilenet_quantized.export
119
  ```
 
120
  ```
121
+ Profiling Results
122
+ ------------------------------------------------------------
123
+ DeepLabV3-Plus-MobileNet-Quantized
124
+ Device : Samsung Galaxy S23 (13)
125
+ Runtime : TFLITE
126
+ Estimated inference time (ms) : 3.3
127
+ Estimated peak memory usage (MB): [0, 146]
128
+ Total # Ops : 104
129
+ Compute Unit(s) : NPU (104 ops)
130
  ```
131
 
132
 
 
165
  Get more details on DeepLabV3-Plus-MobileNet-Quantized's performance across various devices [here](https://aihub.qualcomm.com/models/deeplabv3_plus_mobilenet_quantized).
166
  Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
167
 
168
+
169
  ## License
170
+ * The license for the original implementation of DeepLabV3-Plus-MobileNet-Quantized can be found [here](https://github.com/quic/aimet-model-zoo/blob/develop/LICENSE.pdf).
171
+ * 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)
172
+
173
+
174
 
175
  ## References
176
  * [Rethinking Atrous Convolution for Semantic Image Segmentation](https://arxiv.org/abs/1706.05587)
177
  * [Source Model Implementation](https://github.com/jfzhang95/pytorch-deeplab-xception)
178
 
179
+
180
+
181
  ## Community
182
  * Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
183
  * For questions or feedback please [reach out to us](mailto:[email protected]).