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 2.42, ONNX Runtime 1.24.3	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	Download
QNN_DLC	float	Universal	QAIRT 2.45	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® 8 Elite Gen 5 Mobile	70.44 ms	4 - 327 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® X2 Elite	74.915 ms	53 - 53 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® X Elite	139.411 ms	53 - 53 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Gen 3 Mobile	109.234 ms	1 - 535 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS8550 (Proxy)	147.212 ms	0 - 57 MB	NPU
Unet-Segmentation	ONNX	float	Qualcomm® QCS9075	254.754 ms	9 - 21 MB	NPU
Unet-Segmentation	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	88.283 ms	14 - 331 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	16.467 ms	5 - 190 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® X2 Elite	20.098 ms	29 - 29 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® X Elite	39.086 ms	29 - 29 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	30.356 ms	6 - 340 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS6490	4672.705 ms	935 - 992 MB	CPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	39.83 ms	0 - 2 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCS9075	35.615 ms	4 - 7 MB	NPU
Unet-Segmentation	ONNX	w8a8	Qualcomm® QCM6690	4131.985 ms	838 - 845 MB	CPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	24.658 ms	3 - 188 MB	NPU
Unet-Segmentation	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	3886.044 ms	841 - 848 MB	CPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	67.315 ms	9 - 353 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® X2 Elite	72.071 ms	9 - 9 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® X Elite	132.418 ms	9 - 9 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	102.983 ms	9 - 523 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	953.309 ms	1 - 324 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	134.128 ms	10 - 11 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® SA8775P	240.366 ms	1 - 324 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS9075	248.038 ms	9 - 27 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	279.144 ms	9 - 543 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® SA7255P	953.309 ms	1 - 324 MB	NPU
Unet-Segmentation	QNN_DLC	float	Qualcomm® SA8295P	274.444 ms	0 - 322 MB	NPU
Unet-Segmentation	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	81.482 ms	0 - 331 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	15.685 ms	2 - 197 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® X2 Elite	18.849 ms	2 - 2 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® X Elite	35.682 ms	2 - 2 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	26.203 ms	2 - 320 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS6490	267.959 ms	1 - 7 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	121.478 ms	2 - 181 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	35.048 ms	2 - 4 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® SA8775P	32.181 ms	2 - 181 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS9075	34.587 ms	2 - 8 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCM6690	1170.742 ms	2 - 523 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	53.089 ms	2 - 319 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® SA7255P	121.478 ms	2 - 181 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Qualcomm® SA8295P	63.744 ms	2 - 183 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	21.75 ms	2 - 188 MB	NPU
Unet-Segmentation	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	78.972 ms	2 - 269 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	66.072 ms	6 - 350 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	102.179 ms	6 - 578 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8275 (Proxy)	953.476 ms	1 - 324 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8550 (Proxy)	137.156 ms	6 - 106 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8775P	240.507 ms	6 - 330 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS9075	248.667 ms	0 - 80 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® QCS8450 (Proxy)	278.082 ms	0 - 579 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA7255P	953.476 ms	1 - 324 MB	NPU
Unet-Segmentation	TFLITE	float	Qualcomm® SA8295P	274.486 ms	6 - 328 MB	NPU
Unet-Segmentation	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	82.488 ms	5 - 337 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	15.814 ms	1 - 197 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	26.4 ms	1 - 318 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS6490	267.901 ms	0 - 40 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	121.581 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	34.891 ms	2 - 623 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® SA8775P	32.218 ms	0 - 179 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS9075	34.236 ms	0 - 36 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCM6690	1239.782 ms	0 - 520 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	58.379 ms	2 - 318 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® SA7255P	121.581 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Qualcomm® SA8295P	63.762 ms	2 - 180 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	21.753 ms	2 - 188 MB	NPU
Unet-Segmentation	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	78.581 ms	2 - 267 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.