Yolo-v3: Optimized for Qualcomm Devices

YoloV3 is a machine learning model that predicts bounding boxes and classes of objects in an image.

This is based on the implementation of Yolo-v3 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-v3 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Input resolution: 416p (416x416)
• Model checkpoint: YoloV3 Tiny
• Model size (float): 43.9 MB
• Model size (w8a16): 16.9 MB
• Number of parameters: 11.5M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Yolo-v3	ONNX	float	Snapdragon® X2 Elite	8.243 ms	5 - 5 MB	NPU
Yolo-v3	ONNX	float	Snapdragon® X Elite	4.031 ms	20 - 20 MB	NPU
Yolo-v3	ONNX	float	Snapdragon® 8 Gen 3 Mobile	3.102 ms	0 - 166 MB	NPU
Yolo-v3	ONNX	float	Snapdragon® 8 Gen 1 Mobile	10.551 ms	5 - 194 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	6.791 ms	5 - 8 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.98 ms	0 - 21 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® QCS8450	10.551 ms	5 - 194 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	5.871 ms	5 - 7 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	4.031 ms	20 - 20 MB	NPU
Yolo-v3	ONNX	float	Qualcomm® Dragonwing™ Q-8750	2.54 ms	1 - 152 MB	NPU
Yolo-v3	ONNX	float	Snapdragon® 8 Elite Mobile	2.54 ms	1 - 152 MB	NPU
Yolo-v3	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.899 ms	0 - 152 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® X2 Elite	1.606 ms	2 - 2 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® X Elite	3.829 ms	10 - 10 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	2.558 ms	0 - 89 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	4.518 ms	2 - 88 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	15.525 ms	2 - 5 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-8275	3.747 ms	0 - 4 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.66 ms	0 - 12 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® QCS8450	4.518 ms	2 - 88 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	3.844 ms	2 - 5 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	3.829 ms	10 - 10 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ Q-6690	39.786 ms	0 - 178 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ Q-7790	5.237 ms	2 - 180 MB	NPU
Yolo-v3	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	1.881 ms	0 - 53 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® 8 Elite Mobile	1.881 ms	0 - 53 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.393 ms	0 - 56 MB	NPU
Yolo-v3	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	5.237 ms	2 - 180 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® X2 Elite	2.376 ms	5 - 5 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® X Elite	3.696 ms	5 - 5 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	2.73 ms	5 - 167 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	10.012 ms	5 - 177 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	6.188 ms	5 - 12 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.399 ms	3 - 4 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® SA8775P	5.598 ms	1 - 148 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® SA8650P	5.598 ms	1 - 148 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® SA8255P	5.598 ms	1 - 148 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® QCS8450	10.012 ms	5 - 177 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	5.347 ms	5 - 11 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	3.696 ms	5 - 5 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	2.149 ms	0 - 145 MB	NPU
Yolo-v3	QNN_DLC	float	Qualcomm® SA8295P	9.182 ms	0 - 150 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.149 ms	0 - 145 MB	NPU
Yolo-v3	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.716 ms	5 - 152 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® X2 Elite	4.143 ms	2 - 2 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® X Elite	4.041 ms	2 - 2 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	2.568 ms	2 - 75 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	4.431 ms	2 - 79 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS6490	13.952 ms	2 - 6 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-8275	3.768 ms	1 - 6 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.669 ms	2 - 4 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® SA8775P	4.058 ms	1 - 47 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® SA8650P	4.058 ms	1 - 47 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® SA8255P	4.058 ms	1 - 47 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® QCS8450	4.431 ms	2 - 79 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	3.814 ms	4 - 8 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	4.041 ms	2 - 2 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-6690	37.233 ms	9 - 180 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-7790	4.76 ms	2 - 172 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	1.874 ms	2 - 49 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® SA7255P	9.189 ms	1 - 46 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Qualcomm® SA8295P	5.473 ms	0 - 43 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	1.874 ms	2 - 49 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	1.403 ms	2 - 56 MB	NPU
Yolo-v3	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	4.76 ms	2 - 172 MB	NPU
Yolo-v3	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	2.672 ms	0 - 186 MB	NPU
Yolo-v3	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	9.949 ms	0 - 196 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	6.118 ms	0 - 31 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3.308 ms	0 - 2 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® SA8775P	5.496 ms	0 - 150 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® SA8650P	5.496 ms	0 - 150 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® SA8255P	5.496 ms	0 - 150 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® QCS8450	9.949 ms	0 - 196 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	5.297 ms	0 - 30 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	2.115 ms	0 - 143 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® SA7255P	19.93 ms	0 - 146 MB	NPU
Yolo-v3	TFLITE	float	Qualcomm® SA8295P	9.075 ms	0 - 152 MB	NPU
Yolo-v3	TFLITE	float	Snapdragon® 8 Elite Mobile	2.115 ms	0 - 143 MB	NPU
Yolo-v3	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.596 ms	0 - 149 MB	NPU

License

• The license for the original implementation of Yolo-v3 can be found here.

References

• YOLOv3: An Incremental Improvement
• 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.