Rethinking Atrous Convolution for Semantic Image Segmentation
Paper • 1706.05587 • Published
DeepLabV3-Plus-MobileNet: Optimized for Qualcomm Devices
DeepLabV3 is designed for semantic segmentation at multiple scales, trained on the various datasets. It uses MobileNet as a backbone.
This is based on the implementation of DeepLabV3-Plus-MobileNet found here. This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the Qualcomm® AI Hub Models library to export with custom configurations. More details on model performance across various devices, can be found here.
Qualcomm AI Hub Models uses Qualcomm AI Hub Workbench to compile, profile, and evaluate this model. Sign up to run these models on a hosted Qualcomm® device.
Getting Started
There are two ways to deploy this model on your device:
Option 1: Download Pre-Exported Models
Below are pre-exported model assets ready for deployment.
| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | Download |
| ONNX | w8a16 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | Download |
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | Download |
| QNN_DLC | float | Universal | QAIRT 2.43 | Download |
| QNN_DLC | w8a16 | Universal | QAIRT 2.43 | Download |
| QNN_DLC | w8a8 | Universal | QAIRT 2.43 | Download |
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.19.1 | Download |
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.19.1 | Download |
For more device-specific assets and performance metrics, visit DeepLabV3-Plus-MobileNet on Qualcomm® AI Hub.
Option 2: Export with Custom Configurations
Use the Qualcomm® AI Hub Models Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations
This option is ideal if you need to customize the model beyond the default configuration provided here.
See our repository for DeepLabV3-Plus-MobileNet on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Model checkpoint: VOC2012
- Input resolution: 513x513
- Number of output classes: 21
- Number of parameters: 5.80M
- Model size (float): 22.2 MB
- Model size (w8a16): 6.67 MB
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| DeepLabV3-Plus-MobileNet | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.303 ms | 4 - 178 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Snapdragon® X2 Elite | 5.404 ms | 10 - 10 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Snapdragon® X Elite | 10.986 ms | 10 - 10 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 7.208 ms | 4 - 212 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Qualcomm® QCS8550 (Proxy) | 10.335 ms | 3 - 17 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Qualcomm® QCS9075 | 18.009 ms | 3 - 6 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 5.896 ms | 2 - 172 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 2.964 ms | 0 - 208 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® X2 Elite | 3.71 ms | 7 - 7 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® X Elite | 7.658 ms | 5 - 5 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® 8 Gen 3 Mobile | 5.101 ms | 2 - 220 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Qualcomm® QCS6490 | 1221.703 ms | 89 - 93 MB | CPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Qualcomm® QCS8550 (Proxy) | 7.073 ms | 0 - 8 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Qualcomm® QCS9075 | 8.618 ms | 2 - 4 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Qualcomm® QCM6690 | 588.283 ms | 94 - 102 MB | CPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 3.948 ms | 0 - 182 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a16 | Snapdragon® 7 Gen 4 Mobile | 586.538 ms | 92 - 100 MB | CPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.464 ms | 0 - 179 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Snapdragon® X2 Elite | 1.665 ms | 7 - 7 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Snapdragon® X Elite | 3.903 ms | 5 - 5 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 2.376 ms | 0 - 199 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 3.528 ms | 0 - 98 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Qualcomm® QCS9075 | 4.469 ms | 1 - 4 MB | NPU |
| DeepLabV3-Plus-MobileNet | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.901 ms | 0 - 176 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.292 ms | 3 - 195 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Snapdragon® X2 Elite | 5.56 ms | 3 - 3 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Snapdragon® X Elite | 12.225 ms | 3 - 3 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 8.049 ms | 0 - 212 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 11.505 ms | 3 - 116 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Qualcomm® SA8775P | 17.529 ms | 1 - 171 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Qualcomm® QCS9075 | 19.944 ms | 5 - 10 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 18.951 ms | 2 - 210 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Qualcomm® SA8295P | 19.078 ms | 0 - 174 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.249 ms | 0 - 196 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® 8 Elite Gen 5 Mobile | 3.074 ms | 2 - 197 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® X2 Elite | 4.153 ms | 2 - 2 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® X Elite | 9.185 ms | 2 - 2 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® 8 Gen 3 Mobile | 6.116 ms | 2 - 221 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCS6490 | 33.488 ms | 1 - 4 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCS8275 (Proxy) | 21.924 ms | 2 - 181 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCS8550 (Proxy) | 8.473 ms | 2 - 4 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® SA8775P | 8.963 ms | 2 - 183 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCS9075 | 9.799 ms | 2 - 5 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCM6690 | 106.786 ms | 2 - 232 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® QCS8450 (Proxy) | 12.052 ms | 2 - 220 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® SA7255P | 21.924 ms | 2 - 181 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Qualcomm® SA8295P | 13.743 ms | 2 - 197 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® 8 Elite For Galaxy Mobile | 4.304 ms | 0 - 179 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a16 | Snapdragon® 7 Gen 4 Mobile | 11.992 ms | 2 - 183 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.507 ms | 1 - 175 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 2.053 ms | 1 - 1 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® X Elite | 4.518 ms | 1 - 1 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 2.912 ms | 1 - 195 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 15.472 ms | 3 - 5 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 10.813 ms | 1 - 169 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 4.118 ms | 1 - 2 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® SA8775P | 4.604 ms | 1 - 172 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 5.013 ms | 1 - 3 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 46.422 ms | 1 - 200 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 6.366 ms | 1 - 198 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® SA7255P | 10.813 ms | 1 - 169 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Qualcomm® SA8295P | 6.42 ms | 1 - 167 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 2.143 ms | 1 - 176 MB | NPU |
| DeepLabV3-Plus-MobileNet | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 6.186 ms | 1 - 175 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 4.294 ms | 0 - 194 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.053 ms | 0 - 220 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 58.629 ms | 0 - 170 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 11.454 ms | 0 - 2 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® SA8775P | 17.565 ms | 0 - 174 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® QCS9075 | 19.374 ms | 0 - 18 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 19.088 ms | 0 - 217 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® SA7255P | 58.629 ms | 0 - 170 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Qualcomm® SA8295P | 19.087 ms | 0 - 175 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.261 ms | 0 - 198 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.521 ms | 0 - 182 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 2.939 ms | 0 - 207 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCS6490 | 16.018 ms | 0 - 11 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 10.86 ms | 0 - 173 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 4.329 ms | 0 - 2 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® SA8775P | 4.64 ms | 0 - 175 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCS9075 | 5.181 ms | 0 - 10 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCM6690 | 52.202 ms | 0 - 193 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 5.68 ms | 0 - 204 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® SA7255P | 10.86 ms | 0 - 173 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Qualcomm® SA8295P | 6.24 ms | 0 - 172 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 2.095 ms | 0 - 179 MB | NPU |
| DeepLabV3-Plus-MobileNet | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 6.385 ms | 0 - 178 MB | NPU |
License
- The license for the original implementation of DeepLabV3-Plus-MobileNet can be found here.
References
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.