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:

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

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Yolo-v6	ONNX	float	Snapdragon® X2 Elite	4.356 ms	5 - 5 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® X Elite	9.394 ms	13 - 13 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Gen 3 Mobile	5.927 ms	5 - 226 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Gen 1 Mobile	14.275 ms	5 - 226 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	11.829 ms	5 - 8 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	10.066 ms	0 - 17 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® QCS8450	14.275 ms	5 - 226 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	13.508 ms	5 - 7 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	9.394 ms	13 - 13 MB	NPU
Yolo-v6	ONNX	float	Qualcomm® Dragonwing™ Q-8750	4.608 ms	1 - 186 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Elite Mobile	4.608 ms	1 - 186 MB	NPU
Yolo-v6	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	4.396 ms	0 - 188 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® X2 Elite	4.172 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® X Elite	7.459 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	4.934 ms	0 - 203 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	10.985 ms	4 - 211 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	8.386 ms	5 - 12 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	18.59 ms	2 - 176 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	7.02 ms	5 - 142 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA8775P	8.837 ms	1 - 176 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA8650P	8.837 ms	1 - 176 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA8255P	8.837 ms	1 - 176 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® QCS8450	10.985 ms	4 - 211 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	8.591 ms	7 - 13 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	7.459 ms	5 - 5 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	4.038 ms	5 - 182 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA7255P	18.59 ms	2 - 176 MB	NPU
Yolo-v6	QNN_DLC	float	Qualcomm® SA8295P	9.851 ms	0 - 175 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Elite Mobile	4.038 ms	5 - 182 MB	NPU
Yolo-v6	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	3.783 ms	4 - 183 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® X2 Elite	1.859 ms	2 - 2 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® X Elite	3.942 ms	2 - 2 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	2.229 ms	2 - 91 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	4.936 ms	2 - 92 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	3.364 ms	2 - 7 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	7.051 ms	1 - 64 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.475 ms	2 - 108 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA8775P	4.13 ms	0 - 68 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA8650P	4.13 ms	0 - 68 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA8255P	4.13 ms	0 - 68 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® QCS8450	4.936 ms	2 - 92 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	4.381 ms	0 - 4 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	3.942 ms	2 - 2 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	1.738 ms	2 - 68 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA7255P	7.051 ms	1 - 64 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Qualcomm® SA8295P	5.645 ms	0 - 62 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	1.738 ms	2 - 68 MB	NPU
Yolo-v6	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.38 ms	2 - 73 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	4.958 ms	0 - 214 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	11.028 ms	0 - 214 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	8.282 ms	0 - 25 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	18.533 ms	0 - 183 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	7.037 ms	0 - 3 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8775P	8.772 ms	0 - 185 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8650P	8.772 ms	0 - 185 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8255P	8.772 ms	0 - 185 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® QCS8450	11.028 ms	0 - 214 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	8.498 ms	0 - 24 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	4.076 ms	0 - 184 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA7255P	18.533 ms	0 - 183 MB	NPU
Yolo-v6	TFLITE	float	Qualcomm® SA8295P	9.9 ms	0 - 182 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Elite Mobile	4.076 ms	0 - 184 MB	NPU
Yolo-v6	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	3.777 ms	0 - 187 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.