RTMPose-Body2d: Optimized for Qualcomm Devices

RTMPose is a machine learning model that detects human pose and returns a location and confidence for each of 133 joints.

This is based on the implementation of RTMPose-Body2d 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 2.45, ONNX Runtime 1.27.1	Download
ONNX	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit RTMPose-Body2d 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 RTMPose-Body2d on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.pose_estimation

Model Stats:

• Input resolution: 256x192
• Number of parameters: 17.9M
• Model size (float): 68.5 MB
• Model size (w8a16): 18.2 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
RTMPose-Body2d	ONNX	float	Snapdragon® X2 Elite	0.904 ms	2 - 2 MB	NPU
RTMPose-Body2d	ONNX	float	Snapdragon® X Elite	1.711 ms	37 - 37 MB	NPU
RTMPose-Body2d	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.331 ms	0 - 58 MB	NPU
RTMPose-Body2d	ONNX	float	Snapdragon® 8 Gen 1 Mobile	3.543 ms	0 - 63 MB	NPU
RTMPose-Body2d	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.785 ms	0 - 178 MB	NPU
RTMPose-Body2d	ONNX	float	Qualcomm® QCS8450	3.543 ms	0 - 63 MB	NPU
RTMPose-Body2d	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2.269 ms	0 - 4 MB	NPU
RTMPose-Body2d	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.711 ms	37 - 37 MB	NPU
RTMPose-Body2d	ONNX	float	Qualcomm® Dragonwing™ Q-8750	1.064 ms	0 - 36 MB	NPU
RTMPose-Body2d	ONNX	float	Snapdragon® 8 Elite Mobile	1.064 ms	0 - 36 MB	NPU
RTMPose-Body2d	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	0.935 ms	0 - 38 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® X2 Elite	0.766 ms	1 - 1 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® X Elite	1.739 ms	19 - 19 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	1.14 ms	0 - 87 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	2.185 ms	0 - 90 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® Dragonwing™ QCS6490	5.746 ms	1 - 4 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.678 ms	0 - 140 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® QCS8450	2.185 ms	0 - 90 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-9075	1.853 ms	0 - 3 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® Dragonwing™ IQ-X7181	1.739 ms	19 - 19 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Qualcomm® Dragonwing™ Q-8750	0.856 ms	0 - 63 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® 8 Elite Mobile	0.856 ms	0 - 63 MB	NPU
RTMPose-Body2d	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.734 ms	0 - 59 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® X2 Elite	1.355 ms	1 - 1 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® X Elite	1.843 ms	1 - 1 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.327 ms	0 - 54 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	3.523 ms	0 - 63 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	7.466 ms	1 - 33 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.73 ms	1 - 2 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® SA8775P	2.437 ms	1 - 36 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® SA8650P	2.437 ms	1 - 36 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® SA8255P	2.437 ms	1 - 36 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® QCS8450	3.523 ms	0 - 63 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	2.282 ms	3 - 5 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	1.843 ms	1 - 1 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.083 ms	1 - 35 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® SA7255P	7.466 ms	1 - 33 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Qualcomm® SA8295P	3.524 ms	0 - 35 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.083 ms	1 - 35 MB	NPU
RTMPose-Body2d	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	0.905 ms	1 - 37 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Snapdragon® X2 Elite	0.952 ms	0 - 0 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Snapdragon® X Elite	1.897 ms	0 - 0 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	1.204 ms	0 - 72 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8275	3.812 ms	0 - 48 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.689 ms	0 - 2 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® SA8775P	2.049 ms	0 - 51 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® SA8650P	2.049 ms	0 - 51 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® SA8255P	2.049 ms	0 - 51 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-9075	1.886 ms	0 - 2 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® Dragonwing™ IQ-X7181	1.897 ms	0 - 0 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® Dragonwing™ Q-8750	0.877 ms	0 - 47 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Qualcomm® SA7255P	3.812 ms	0 - 48 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	0.877 ms	0 - 47 MB	NPU
RTMPose-Body2d	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	0.74 ms	0 - 51 MB	NPU
RTMPose-Body2d	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.257 ms	0 - 78 MB	NPU
RTMPose-Body2d	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	3.506 ms	0 - 85 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	7.473 ms	0 - 44 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.711 ms	0 - 2 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® SA8775P	2.378 ms	0 - 48 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® SA8650P	2.378 ms	0 - 48 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® SA8255P	2.378 ms	0 - 48 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® QCS8450	3.506 ms	0 - 85 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	2.22 ms	0 - 40 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.037 ms	0 - 44 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® SA7255P	7.473 ms	0 - 44 MB	NPU
RTMPose-Body2d	TFLITE	float	Qualcomm® SA8295P	3.51 ms	0 - 47 MB	NPU
RTMPose-Body2d	TFLITE	float	Snapdragon® 8 Elite Mobile	1.037 ms	0 - 44 MB	NPU
RTMPose-Body2d	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	0.904 ms	0 - 46 MB	NPU

License

• The license for the original implementation of RTMPose-Body2d can be found here.

References

• RTMPose: Real-Time Multi-Person Pose Estimation based on MMPose
• 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.