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® X2 Elite | 74.872 ms | 16 - 16 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® X Elite | 141.902 ms | 54 - 54 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 115.045 ms | 24 - 546 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Gen 1 Mobile | 282.092 ms | 1 - 548 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-8275 | 285.991 ms | 9 - 22 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 153.301 ms | 0 - 57 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® QCS8450 | 282.092 ms | 1 - 548 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-9075 | 250.636 ms | 9 - 22 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ IQ-X7181 | 141.902 ms | 54 - 54 MB | NPU |
| Unet-Segmentation | ONNX | float | Qualcomm® Dragonwing™ Q-8750 | 92.842 ms | 15 - 333 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite Mobile | 92.842 ms | 15 - 333 MB | NPU |
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 65.231 ms | 14 - 343 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X2 Elite | 18.908 ms | 5 - 5 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X Elite | 37.626 ms | 29 - 29 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 29.867 ms | 6 - 338 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Gen 1 Mobile | 62.637 ms | 6 - 341 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 299.62 ms | 3 - 8 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 40.828 ms | 4 - 7 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 38.12 ms | 0 - 32 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCS8450 | 62.637 ms | 6 - 341 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 35.132 ms | 4 - 7 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 37.626 ms | 29 - 29 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1231.467 ms | 0 - 542 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 81.435 ms | 8 - 282 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 24.358 ms | 3 - 190 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite Mobile | 24.358 ms | 3 - 190 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 16.714 ms | 3 - 190 MB | NPU |
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 81.435 ms | 8 - 282 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 71.738 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 132.584 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 102.25 ms | 9 - 526 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 1 Mobile | 273.152 ms | 9 - 545 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-8275 | 274.137 ms | 9 - 27 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 137.703 ms | 10 - 19 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 240.377 ms | 1 - 325 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8650P | 240.377 ms | 1 - 325 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8255P | 240.377 ms | 1 - 325 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 | 273.152 ms | 9 - 545 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-9075 | 239.327 ms | 9 - 27 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ IQ-X7181 | 132.584 ms | 9 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® Dragonwing™ Q-8750 | 82.577 ms | 0 - 331 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 953.29 ms | 0 - 324 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 274.483 ms | 0 - 323 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Mobile | 82.577 ms | 0 - 331 MB | NPU |
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 61.334 ms | 3 - 348 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 27.399 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X Elite | 35.741 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.305 ms | 2 - 318 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Gen 1 Mobile | 54.11 ms | 2 - 317 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 288.689 ms | 2 - 8 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 37.219 ms | 2 - 9 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 34.613 ms | 2 - 5 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8775P | 32.179 ms | 1 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8650P | 32.179 ms | 1 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8255P | 32.179 ms | 1 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS8450 | 54.11 ms | 2 - 317 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 32.609 ms | 1 - 7 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ IQ-X7181 | 35.741 ms | 2 - 2 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1200.713 ms | 2 - 525 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 78.744 ms | 2 - 273 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 21.682 ms | 2 - 187 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA7255P | 121.489 ms | 2 - 181 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8295P | 63.732 ms | 0 - 179 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite Mobile | 21.682 ms | 2 - 187 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 15.72 ms | 2 - 201 MB | NPU |
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 78.744 ms | 2 - 273 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 103.236 ms | 6 - 574 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Gen 1 Mobile | 276.654 ms | 7 - 590 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-8275 | 274.063 ms | 0 - 81 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 141.99 ms | 7 - 218 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8775P | 240.507 ms | 7 - 331 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8650P | 240.507 ms | 7 - 331 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8255P | 240.507 ms | 7 - 331 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS8450 | 276.654 ms | 7 - 590 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ IQ-9075 | 238.517 ms | 0 - 80 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® Dragonwing™ Q-8750 | 82.2 ms | 6 - 336 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA7255P | 953.454 ms | 5 - 328 MB | NPU |
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8295P | 274.486 ms | 7 - 329 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite Mobile | 82.2 ms | 6 - 336 MB | NPU |
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 61.864 ms | 6 - 354 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.181 ms | 1 - 317 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Gen 1 Mobile | 58.358 ms | 2 - 318 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS6490 | 288.847 ms | 0 - 41 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-8275 | 37.257 ms | 2 - 39 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ QCS8550 (Proxy) | 32.487 ms | 1 - 4 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8775P | 32.211 ms | 2 - 180 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8650P | 32.211 ms | 2 - 180 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8255P | 32.211 ms | 2 - 180 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS8450 | 58.358 ms | 2 - 318 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ IQ-9075 | 32.585 ms | 1 - 38 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-6690 | 1175.945 ms | 0 - 522 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-7790 | 79.045 ms | 0 - 264 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® Dragonwing™ Q-8750 | 21.803 ms | 1 - 187 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA7255P | 121.59 ms | 2 - 181 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8295P | 63.769 ms | 2 - 180 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite Mobile | 21.803 ms | 1 - 187 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 16.058 ms | 2 - 198 MB | NPU |
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 79.045 ms | 0 - 264 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.
