Unet-Segmentation: Optimized for Qualcomm Devices
UNet is a machine learning model that produces a segmentation mask for an image. The most basic use case will label each pixel in the image as being in the foreground or the background. More advanced usage will assign a class label to each pixel. This version of the model was trained on the data from Kaggle's Carvana Image Masking Challenge (see https://www.kaggle.com/c/carvana-image-masking-challenge) and is used for vehicle segmentation.
This is based on the implementation of Unet-Segmentation 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
Due to licensing restrictions, we cannot distribute pre-exported model assets for this model. 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
See our repository for Unet-Segmentation on GitHub for usage instructions.
Model Details
Model Type: Model_use_case.semantic_segmentation
Model Stats:
- Input resolution: 640x1280
- Model checkpoint: unet_carvana_scale1.0_epoch2
- Model size (float): 118 MB
- Model size (w8a8): 29.8 MB
- Number of output classes: 2 (foreground / background)
- Number of parameters: 31.0M
Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit |
|---|---|---|---|---|---|---|
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 59.005 ms | 14 - 280 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 81.346 ms | 16 - 266 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® X2 Elite | 71.104 ms | 17 - 17 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® X Elite | 132.33 ms | 54 - 54 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 101.666 ms | 3 - 455 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 262.01 ms | 1 - 479 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 274.151 ms | 9 - 23 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 137.295 ms | 0 - 56 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® QCS8450 | 262.01 ms | 1 - 479 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 239.75 ms | 9 - 22 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 132.33 ms | 54 - 54 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 81.346 ms | 16 - 266 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 16.264 ms | 3 - 182 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 21.606 ms | 3 - 180 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X2 Elite | 18.654 ms | 4 - 4 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X Elite | 36.501 ms | 29 - 29 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.128 ms | 6 - 328 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 48.099 ms | 6 - 328 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 246.461 ms | 3 - 9 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 37.005 ms | 4 - 8 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 37.234 ms | 0 - 32 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCS8450 | 48.099 ms | 6 - 328 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 32.156 ms | 4 - 7 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 36.501 ms | 29 - 29 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1041.701 ms | 0 - 497 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 74.51 ms | 6 - 261 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 21.606 ms | 3 - 180 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 74.51 ms | 6 - 261 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 54.526 ms | 0 - 276 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 75.423 ms | 0 - 262 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 68.019 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 125.744 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 102.472 ms | 9 - 461 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 237.965 ms | 4 - 468 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 265.804 ms | 9 - 28 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 129.143 ms | 10 - 47 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 233.605 ms | 1 - 255 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8650P | 233.605 ms | 1 - 255 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8255P | 233.605 ms | 1 - 255 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 | 237.965 ms | 4 - 468 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 232.275 ms | 9 - 27 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 125.744 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 75.423 ms | 0 - 262 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 944.38 ms | 0 - 254 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 253.721 ms | 0 - 254 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 15.328 ms | 2 - 188 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 21.492 ms | 2 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 18.29 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X Elite | 34.725 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 25.242 ms | 2 - 312 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 44.42 ms | 2 - 310 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 243.355 ms | 2 - 8 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 35.888 ms | 2 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 33.184 ms | 3 - 239 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 44.42 ms | 2 - 310 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 31.257 ms | 1 - 7 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 34.725 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1015.46 ms | 2 - 475 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 70.699 ms | 2 - 250 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 21.492 ms | 2 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA7255P | 119.629 ms | 0 - 169 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8295P | 62.119 ms | 0 - 173 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 70.699 ms | 2 - 250 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 54.465 ms | 6 - 282 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 76.042 ms | 5 - 266 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 102.833 ms | 6 - 511 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 237.527 ms | 7 - 516 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 265.595 ms | 0 - 81 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 126.85 ms | 7 - 443 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8775P | 233.636 ms | 7 - 260 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8650P | 233.636 ms | 7 - 260 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8255P | 233.636 ms | 7 - 260 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS8450 | 237.527 ms | 7 - 516 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 232.413 ms | 0 - 80 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 76.042 ms | 5 - 266 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA7255P | 944.688 ms | 7 - 260 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8295P | 253.777 ms | 0 - 252 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 For Galaxy Mobile | 15.345 ms | 0 - 188 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 21.406 ms | 1 - 178 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 25.044 ms | 1 - 308 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 48.362 ms | 2 - 311 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 243.405 ms | 0 - 42 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 36.112 ms | 2 - 39 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 30.913 ms | 2 - 4 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8775P | 30.635 ms | 2 - 172 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8650P | 30.635 ms | 2 - 172 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8255P | 30.635 ms | 2 - 172 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS8450 | 48.362 ms | 2 - 311 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 31.101 ms | 2 - 39 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1048.638 ms | 0 - 472 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 71.248 ms | 1 - 244 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 21.406 ms | 1 - 178 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8295P | 62.151 ms | 2 - 171 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 71.248 ms | 1 - 244 MB | NPU |
License
- The license for the original implementation of Unet-Segmentation 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.
