RF-DETR: Optimized for Qualcomm Devices

DETR is a machine learning model that can detect objects (trained on COCO dataset).

This is based on the implementation of RF-DETR 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
QNN_DLC	float	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Model checkpoint: RF-DETR-small
• Input resolution: 512x512
• Supported variants: nano (384x384), small (512x512), medium (576x576), base (560x560)
• Number of parameters: 28.5M
• Model size (float): 109 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
RF-DETR	ONNX	float	Snapdragon® 8 Gen 3 Mobile	30.957 ms	0 - 408 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Gen 1 Mobile	99.441 ms	0 - 418 MB	NPU
RF-DETR	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	41.778 ms	0 - 65 MB	NPU
RF-DETR	ONNX	float	Qualcomm® QCS8450	99.441 ms	0 - 418 MB	NPU
RF-DETR	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	50.983 ms	10 - 16 MB	NPU
RF-DETR	ONNX	float	Qualcomm® Dragonwing™ Q-8750	23.56 ms	13 - 332 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Elite Mobile	23.56 ms	13 - 332 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	20.168 ms	10 - 365 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® X2 Elite	25.076 ms	3 - 3 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® X Elite	50.836 ms	3 - 3 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	36.389 ms	0 - 469 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	99.357 ms	0 - 472 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	135.413 ms	1 - 359 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	50.294 ms	3 - 6 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8775P	56.797 ms	1 - 383 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8650P	56.797 ms	1 - 383 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8255P	56.797 ms	1 - 383 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® QCS8450	99.357 ms	0 - 472 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	68.479 ms	3 - 8 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	50.836 ms	3 - 3 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	28.838 ms	0 - 409 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA7255P	135.413 ms	1 - 359 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8295P	81.298 ms	0 - 360 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Elite Mobile	28.838 ms	0 - 409 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	23.55 ms	3 - 406 MB	NPU

License

• The license for the original implementation of RF-DETR can be found here.

References

• RF-DETR A SOTA Real-Time Object Detection Model
• 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.