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

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 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Unet-Segmentation 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 Unet-Segmentation on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.semantic_segmentation

Model Stats:

• Model checkpoint: unet_carvana_scale1.0_epoch2
• Input resolution: 640x1280
• 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® X2 Elite	74.947 ms	171 - 171 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® X Elite	142.574 ms	173 - 173 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Gen 3 Mobile	113.477 ms	24 - 534 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Gen 1 Mobile	279.768 ms	10 - 547 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS8550 (Proxy)	149.803 ms	0 - 62 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS8450	279.768 ms	10 - 547 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Elite Mobile	90.992 ms	15 - 334 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	66.353 ms	13 - 339 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS9075	254.674 ms	9 - 64 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS8750	90.992 ms	15 - 334 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS7181	142.574 ms	173 - 173 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® X2 Elite	18.747 ms	210 - 210 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® X Elite	37.643 ms	179 - 179 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	29.966 ms	0 - 323 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	66.58 ms	0 - 325 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS6490	271.558 ms	4 - 49 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	38.652 ms	0 - 33 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS8450	66.58 ms	0 - 325 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	16.442 ms	3 - 187 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	81.197 ms	6 - 282 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCM6690	1252.373 ms	0 - 534 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS9075	35.626 ms	4 - 49 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Elite Mobile	24.504 ms	3 - 190 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS7790	81.197 ms	6 - 282 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS8750	24.504 ms	3 - 190 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS7181	37.643 ms	179 - 179 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® X2 Elite	71.761 ms	9 - 9 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® X Elite	132.352 ms	9 - 9 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	102.412 ms	9 - 524 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	272.973 ms	4 - 539 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8275	953.516 ms	1 - 324 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	137.077 ms	10 - 11 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8450	272.973 ms	4 - 539 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Elite Mobile	82.191 ms	9 - 341 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® SA8295P	274.442 ms	0 - 323 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	61.276 ms	9 - 356 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® SA7255P	953.516 ms	1 - 324 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS9075	248.054 ms	9 - 27 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8750	82.191 ms	9 - 341 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS7181	132.352 ms	9 - 9 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® X2 Elite	18.875 ms	2 - 2 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® X Elite	35.775 ms	2 - 2 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	26.305 ms	2 - 319 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	59.676 ms	3 - 317 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS6490	267.25 ms	4 - 10 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8275	121.518 ms	1 - 180 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	34.855 ms	3 - 6 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8450	59.676 ms	3 - 317 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	15.692 ms	2 - 200 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	78.865 ms	2 - 272 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCM6690	1211.427 ms	3 - 524 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS9075	35.056 ms	2 - 8 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® SA7255P	121.518 ms	1 - 180 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	22.027 ms	2 - 188 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® SA8295P	63.758 ms	0 - 180 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS7790	78.865 ms	2 - 272 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8750	22.027 ms	2 - 188 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS7181	35.775 ms	2 - 2 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	105.463 ms	6 - 577 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	280.367 ms	7 - 589 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8275	953.647 ms	1 - 325 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8550 (Proxy)	138.416 ms	6 - 443 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8775P	2218.877 ms	3 - 21 MB	GPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8650P	2218.877 ms	3 - 21 MB	GPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8255P	2218.877 ms	3 - 21 MB	GPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8450	280.367 ms	7 - 589 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Elite Mobile	82.53 ms	5 - 335 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8295P	274.486 ms	7 - 329 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	61.496 ms	6 - 352 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA7255P	953.647 ms	1 - 325 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS9075	248.287 ms	0 - 80 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8750	82.53 ms	5 - 335 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	26.279 ms	1 - 317 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	60.395 ms	2 - 317 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS6490	267.778 ms	2 - 41 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8275	121.58 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	32.11 ms	2 - 4 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8450	60.395 ms	2 - 317 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	15.745 ms	2 - 199 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	78.635 ms	1 - 266 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCM6690	1211.941 ms	0 - 522 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS9075	34.237 ms	1 - 38 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® SA7255P	121.58 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	21.725 ms	2 - 188 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® SA8295P	63.758 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS7790	78.635 ms	1 - 266 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8750	21.725 ms	2 - 188 MB	NPU

License

• The license for the original implementation of Unet-Segmentation can be found here.

References

• U-Net: Convolutional Networks for Biomedical Image Segmentation
• Source Model Implementation

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.