MediaPipe-Hand-Detection: Optimized for Qualcomm Devices

The MediaPipe Hand Landmark Detector is a machine learning pipeline that predicts bounding boxes and pose skeletons of hands in an image.

This is based on the implementation of MediaPipe-Hand-Detection 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 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit MediaPipe-Hand-Detection 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 MediaPipe-Hand-Detection on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Input resolution: 256x256
• Number of parameters (hand_detector): 1.76M
• Model size (hand_detector) (float): 6.75 MB
• Number of parameters (hand_landmark_detector): 2.01M
• Model size (hand_landmark_detector) (float): 7.70 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
hand_detector	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.417 ms	0 - 46 MB	NPU
hand_detector	ONNX	float	Snapdragon® X2 Elite	0.489 ms	212 - 212 MB	NPU
hand_detector	ONNX	float	Snapdragon® X Elite	0.864 ms	148 - 148 MB	NPU
hand_detector	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.593 ms	0 - 66 MB	NPU
hand_detector	ONNX	float	Qualcomm® QCS8550 (Proxy)	0.847 ms	0 - 14 MB	NPU
hand_detector	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.497 ms	0 - 42 MB	NPU
hand_detector	ONNX	float	Qualcomm® QCS9075	1.275 ms	0 - 49 MB	NPU
hand_detector	ONNX	float	Qualcomm® QCS8750	0.497 ms	0 - 42 MB	NPU
hand_detector	ONNX	float	Qualcomm® QCS7181	0.864 ms	148 - 148 MB	NPU
hand_detector	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.378 ms	1 - 42 MB	NPU
hand_detector	QNN_DLC	float	Snapdragon® X2 Elite	0.602 ms	1 - 1 MB	NPU
hand_detector	QNN_DLC	float	Snapdragon® X Elite	0.877 ms	1 - 1 MB	NPU
hand_detector	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.531 ms	0 - 57 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	0.729 ms	1 - 2 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® SA8775P	1.271 ms	1 - 42 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® SA8650P	1.271 ms	1 - 42 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® SA8255P	1.271 ms	1 - 42 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	1.378 ms	0 - 53 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® SA8295P	1.705 ms	1 - 32 MB	NPU
hand_detector	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.441 ms	0 - 39 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® QCS9075	1.116 ms	1 - 3 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® QCS8750	0.441 ms	0 - 39 MB	NPU
hand_detector	QNN_DLC	float	Qualcomm® QCS7181	0.877 ms	1 - 1 MB	NPU
hand_detector	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.379 ms	0 - 42 MB	NPU
hand_detector	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.535 ms	0 - 56 MB	NPU
hand_detector	TFLITE	float	Qualcomm® QCS8550 (Proxy)	0.727 ms	0 - 2 MB	NPU
hand_detector	TFLITE	float	Qualcomm® SA8775P	1.293 ms	0 - 41 MB	NPU
hand_detector	TFLITE	float	Qualcomm® SA8650P	1.293 ms	0 - 41 MB	NPU
hand_detector	TFLITE	float	Qualcomm® SA8255P	1.293 ms	0 - 41 MB	NPU
hand_detector	TFLITE	float	Qualcomm® QCS8450 (Proxy)	1.384 ms	0 - 50 MB	NPU
hand_detector	TFLITE	float	Qualcomm® SA8295P	1.714 ms	0 - 30 MB	NPU
hand_detector	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.447 ms	0 - 41 MB	NPU
hand_detector	TFLITE	float	Qualcomm® QCS9075	1.133 ms	0 - 7 MB	NPU
hand_detector	TFLITE	float	Qualcomm® QCS8750	0.447 ms	0 - 41 MB	NPU
hand_landmark_detector	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.558 ms	0 - 42 MB	NPU
hand_landmark_detector	ONNX	float	Snapdragon® X2 Elite	0.718 ms	211 - 211 MB	NPU
hand_landmark_detector	ONNX	float	Snapdragon® X Elite	1.152 ms	180 - 180 MB	NPU
hand_landmark_detector	ONNX	float	Snapdragon® 8 Gen 3 Mobile	0.83 ms	0 - 67 MB	NPU
hand_landmark_detector	ONNX	float	Qualcomm® QCS8550 (Proxy)	1.101 ms	0 - 58 MB	NPU
hand_landmark_detector	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	0.658 ms	0 - 38 MB	NPU
hand_landmark_detector	ONNX	float	Qualcomm® QCS9075	1.827 ms	0 - 51 MB	NPU
hand_landmark_detector	ONNX	float	Qualcomm® QCS8750	0.658 ms	0 - 38 MB	NPU
hand_landmark_detector	ONNX	float	Qualcomm® QCS7181	1.152 ms	180 - 180 MB	NPU
hand_landmark_detector	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.511 ms	1 - 39 MB	NPU
hand_landmark_detector	QNN_DLC	float	Snapdragon® X2 Elite	0.838 ms	1 - 1 MB	NPU
hand_landmark_detector	QNN_DLC	float	Snapdragon® X Elite	1.278 ms	1 - 1 MB	NPU
hand_landmark_detector	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	0.765 ms	0 - 56 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	1.08 ms	1 - 75 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® SA8775P	1.872 ms	1 - 37 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® SA8650P	1.872 ms	1 - 37 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® SA8255P	1.872 ms	1 - 37 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	1.954 ms	0 - 56 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® SA8295P	2.236 ms	0 - 30 MB	NPU
hand_landmark_detector	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	0.599 ms	0 - 38 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® QCS9075	1.669 ms	1 - 3 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® QCS8750	0.599 ms	0 - 38 MB	NPU
hand_landmark_detector	QNN_DLC	float	Qualcomm® QCS7181	1.278 ms	1 - 1 MB	NPU
hand_landmark_detector	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.51 ms	0 - 43 MB	NPU
hand_landmark_detector	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	0.762 ms	0 - 58 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® QCS8550 (Proxy)	0.992 ms	0 - 2 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® SA8775P	1.85 ms	0 - 43 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® SA8650P	1.85 ms	0 - 43 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® SA8255P	1.85 ms	0 - 43 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® QCS8450 (Proxy)	1.96 ms	0 - 60 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® SA8295P	2.242 ms	0 - 34 MB	NPU
hand_landmark_detector	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	0.604 ms	0 - 43 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® QCS9075	1.681 ms	0 - 9 MB	NPU
hand_landmark_detector	TFLITE	float	Qualcomm® QCS8750	0.604 ms	0 - 43 MB	NPU

License

• The license for the original implementation of MediaPipe-Hand-Detection can be found here.

References

• MediaPipe Hands: On-device Real-time Hand Tracking
• 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.