---
library_name: pytorch
license: other
tags:
- backbone
- bu_auto
- real_time
- android
pipeline_tag: image-segmentation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/web-assets/model_demo.png)

# 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](https://github.com/milesial/Pytorch-UNet).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/unet_segmentation) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) 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 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.42, ONNX Runtime 1.24.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.43 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.43, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/unet_segmentation/releases/v0.47.0/unet_segmentation-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[Unet-Segmentation on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/unet_segmentation)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/unet_segmentation) 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 [Unet-Segmentation on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/unet_segmentation) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: unet_carvana_scale1.0_epoch2
- Input resolution: 224x224
- Number of output classes: 2 (foreground / background)
- Number of parameters: 31.0M
- Model size (float): 118 MB
- Model size (w8a8): 29.8 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| Unet-Segmentation | ONNX | float | Snapdragon® X Elite | 139.479 ms | 53 - 53 MB | NPU
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 110.033 ms | 2 - 538 MB | NPU
| Unet-Segmentation | ONNX | float | Qualcomm® QCS8550 (Proxy) | 149.746 ms | 0 - 57 MB | NPU
| Unet-Segmentation | ONNX | float | Qualcomm® QCS9075 | 254.819 ms | 9 - 21 MB | NPU
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 88.65 ms | 15 - 330 MB | NPU
| Unet-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 70.141 ms | 24 - 344 MB | NPU
| Unet-Segmentation | ONNX | float | Snapdragon® X2 Elite | 75.136 ms | 53 - 53 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X Elite | 39.056 ms | 29 - 29 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 30.222 ms | 6 - 338 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCS6490 | 4647.573 ms | 942 - 1000 MB | CPU
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 41.711 ms | 4 - 6 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCS9075 | 35.615 ms | 4 - 7 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Qualcomm® QCM6690 | 4142.982 ms | 836 - 842 MB | CPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 24.867 ms | 3 - 191 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 3886.757 ms | 834 - 842 MB | CPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 16.251 ms | 0 - 185 MB | NPU
| Unet-Segmentation | ONNX | w8a8 | Snapdragon® X2 Elite | 20.144 ms | 29 - 29 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 132.389 ms | 9 - 9 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 101.504 ms | 9 - 547 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 953.65 ms | 0 - 322 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 135.421 ms | 10 - 13 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 240.476 ms | 0 - 324 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS9075 | 248.263 ms | 9 - 27 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 274.543 ms | 7 - 549 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 953.65 ms | 0 - 322 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 274.502 ms | 0 - 322 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 82.334 ms | 0 - 331 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 65.76 ms | 9 - 350 MB | NPU
| Unet-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 71.586 ms | 9 - 9 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X Elite | 35.715 ms | 2 - 2 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.079 ms | 2 - 321 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 267.761 ms | 2 - 8 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 35.233 ms | 2 - 35 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8775P | 32.227 ms | 1 - 180 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 34.981 ms | 2 - 8 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 1232.21 ms | 3 - 522 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 57.992 ms | 2 - 320 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Qualcomm® SA8295P | 63.705 ms | 0 - 180 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 21.942 ms | 2 - 191 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 78.791 ms | 2 - 268 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 15.788 ms | 2 - 197 MB | NPU
| Unet-Segmentation | QNN_DLC | w8a8 | Snapdragon® X2 Elite | 18.916 ms | 2 - 2 MB | NPU
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 102.428 ms | 0 - 535 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 953.609 ms | 0 - 325 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 156.101 ms | 6 - 443 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8775P | 240.55 ms | 6 - 331 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS9075 | 248.108 ms | 0 - 80 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 277.793 ms | 7 - 552 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® SA7255P | 953.609 ms | 0 - 325 MB | NPU
| Unet-Segmentation | TFLITE | float | Qualcomm® SA8295P | 274.559 ms | 6 - 329 MB | NPU
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 81.817 ms | 4 - 337 MB | NPU
| Unet-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 67.608 ms | 5 - 351 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 26.281 ms | 0 - 318 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS6490 | 267.819 ms | 0 - 40 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 121.596 ms | 2 - 181 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 34.978 ms | 0 - 651 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8775P | 32.263 ms | 2 - 181 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS9075 | 34.225 ms | 1 - 38 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCM6690 | 1238.794 ms | 0 - 519 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 58.357 ms | 2 - 321 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA7255P | 121.596 ms | 2 - 181 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Qualcomm® SA8295P | 63.788 ms | 2 - 180 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 21.845 ms | 1 - 187 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 78.641 ms | 2 - 266 MB | NPU
| Unet-Segmentation | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 15.774 ms | 7 - 202 MB | NPU

## License
* The license for the original implementation of Unet-Segmentation can be found
  [here](https://github.com/milesial/Pytorch-UNet/blob/master/LICENSE).

## References
* [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://arxiv.org/abs/1505.04597)
* [Source Model Implementation](https://github.com/milesial/Pytorch-UNet)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).