Yolo-v6: Optimized for Qualcomm Devices

YoloV6 is a machine learning model that predicts bounding boxes and classes of objects in an image.

This is based on the implementation of Yolo-v6 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 Yolo-v6 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Model checkpoint: YoloV6-N
• Input resolution: 640x640
• Number of parameters: 4.68M
• Model size (float): 17.9 MB
• Model size (w8a8): 4.68 MB
• Model size (w8a16): 5.03 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Yolo-v6	ONNX	float	Snapdragon® X2 Elite	4.016 ms	208 - 208 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® X Elite	9.296 ms	177 - 177 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Gen 3 Mobile	7.855 ms	5 - 224 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Gen 1 Mobile	13.965 ms	5 - 224 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS8550 (Proxy)	10.028 ms	5 - 9 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS8450	13.965 ms	5 - 224 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Elite Mobile	4.561 ms	1 - 188 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	4.297 ms	3 - 190 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS9075	10.354 ms	5 - 50 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS8750	4.561 ms	1 - 188 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS7181	9.296 ms	177 - 177 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® X2 Elite	4.06 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® X Elite	7.194 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	4.884 ms	0 - 203 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	11.181 ms	5 - 212 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS8275	18.336 ms	0 - 174 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	6.818 ms	5 - 7 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS8450	11.181 ms	5 - 212 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.241 ms	0 - 179 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA7255P	18.336 ms	0 - 174 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA8295P	9.922 ms	0 - 175 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.752 ms	5 - 184 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS9075	8.782 ms	7 - 13 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS8750	4.241 ms	0 - 179 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS7181	7.194 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® X2 Elite	1.891 ms	2 - 2 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® X Elite	3.984 ms	2 - 2 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	2.278 ms	2 - 92 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	4.864 ms	2 - 91 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS8275	7.215 ms	0 - 64 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	3.525 ms	2 - 4 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS8450	4.864 ms	2 - 91 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.362 ms	2 - 73 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	1.731 ms	2 - 70 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS9075	4.169 ms	2 - 6 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA8295P	5.604 ms	0 - 62 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA7255P	7.215 ms	0 - 64 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS8750	1.731 ms	2 - 70 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS7181	3.984 ms	2 - 2 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	4.91 ms	0 - 213 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	10.906 ms	0 - 217 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® QCS8275	18.353 ms	0 - 184 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® QCS8550 (Proxy)	6.797 ms	0 - 3 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8775P	29.156 ms	6 - 38 MB	GPU
Yolo-v6	TFLITE	float	Qualcomm® SA8650P	29.156 ms	6 - 38 MB	GPU
Yolo-v6	TFLITE	float	Qualcomm® SA8255P	29.156 ms	6 - 38 MB	GPU
Yolo-v6	TFLITE	float	Qualcomm® QCS8450	10.906 ms	0 - 217 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Elite Mobile	4.174 ms	0 - 190 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA7255P	18.353 ms	0 - 184 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8295P	9.992 ms	0 - 183 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.768 ms	0 - 187 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® QCS9075	9.096 ms	0 - 20 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® QCS8750	4.174 ms	0 - 190 MB	NPU

License

• The license for the original implementation of Yolo-v6 can be found here.

References

• YOLOv6: A Single-Stage Object Detection Framework for Industrial Applications
• 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.