Yolo-v3: Optimized for Mobile Deployment

Real-time object detection optimized for mobile and edge

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

This model is an implementation of Yolo-v3 found here.

More details on model performance across various devices, can be found here.

Model Details

• Model Type: Object detection
• Model Stats:
  • Model checkpoint: YoloV3 Tiny
  • Input resolution: 416p (416x416)
  • Number of parameters: 8.85M
  • Model size: 24.4 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
Yolo-v3	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	16.507 ms	0 - 12 MB	FP16	NPU	--
Yolo-v3	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	8.225 ms	5 - 8 MB	FP16	NPU	--
Yolo-v3	Samsung Galaxy S23	Snapdragon® 8 Gen 2	ONNX	12.251 ms	6 - 76 MB	FP16	NPU	--
Yolo-v3	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	10.908 ms	0 - 94 MB	FP16	NPU	--
Yolo-v3	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	6.055 ms	5 - 23 MB	FP16	NPU	--
Yolo-v3	Samsung Galaxy S24	Snapdragon® 8 Gen 3	ONNX	9.099 ms	11 - 41 MB	FP16	NPU	--
Yolo-v3	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	8.415 ms	0 - 73 MB	FP16	NPU	--
Yolo-v3	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	7.0 ms	5 - 32 MB	FP16	NPU	--
Yolo-v3	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	ONNX	10.918 ms	11 - 35 MB	FP16	NPU	--
Yolo-v3	QCS8275 (Proxy)	QCS8275 Proxy	TFLITE	107.795 ms	0 - 70 MB	FP16	NPU	--
Yolo-v3	QCS8275 (Proxy)	QCS8275 Proxy	QNN	97.923 ms	1 - 9 MB	FP16	NPU	--
Yolo-v3	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	16.743 ms	0 - 12 MB	FP16	NPU	--
Yolo-v3	QCS8550 (Proxy)	QCS8550 Proxy	QNN	8.097 ms	5 - 7 MB	FP16	NPU	--
Yolo-v3	QCS9075 (Proxy)	QCS9075 Proxy	TFLITE	21.161 ms	0 - 71 MB	FP16	NPU	--
Yolo-v3	QCS9075 (Proxy)	QCS9075 Proxy	QNN	11.588 ms	0 - 10 MB	FP16	NPU	--
Yolo-v3	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	17.938 ms	0 - 88 MB	FP16	NPU	--
Yolo-v3	QCS8450 (Proxy)	QCS8450 Proxy	QNN	12.698 ms	5 - 27 MB	FP16	NPU	--
Yolo-v3	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	8.438 ms	5 - 5 MB	FP16	NPU	--
Yolo-v3	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	13.752 ms	5 - 5 MB	FP16	NPU	--

License

• The license for the original implementation of Yolo-v3 can be found here.
• The license for the compiled assets for on-device deployment 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.

Usage and Limitations

Model may not be used for or in connection with any of the following applications:

• Accessing essential private and public services and benefits;
• Administration of justice and democratic processes;
• Assessing or recognizing the emotional state of a person;
• Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics;
• Education and vocational training;
• Employment and workers management;
• Exploitation of the vulnerabilities of persons resulting in harmful behavior;
• General purpose social scoring;
• Law enforcement;
• Management and operation of critical infrastructure;
• Migration, asylum and border control management;
• Predictive policing;
• Real-time remote biometric identification in public spaces;
• Recommender systems of social media platforms;
• Scraping of facial images (from the internet or otherwise); and/or
• Subliminal manipulation