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 2.42, ONNX Runtime 1.24.3	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 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-base
• Input resolution: 560x560
• Number of parameters: 29.0M
• Model size: 116MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
RF-DETR	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	30.939 ms	16 - 795 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Elite Mobile	35.758 ms	14 - 767 MB	NPU
RF-DETR	ONNX	float	Snapdragon® X2 Elite	28.315 ms	55 - 55 MB	NPU
RF-DETR	ONNX	float	Snapdragon® X Elite	63.12 ms	53 - 53 MB	NPU
RF-DETR	ONNX	float	Snapdragon® X Elite	63.12 ms	53 - 53 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Gen 3 Mobile	47.607 ms	2 - 910 MB	NPU
RF-DETR	ONNX	float	Qualcomm® QCS8550 (Proxy)	61.35 ms	13 - 21 MB	NPU
RF-DETR	ONNX	float	Qualcomm® QCS9075	79.562 ms	13 - 20 MB	NPU
RF-DETR	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	35.758 ms	14 - 767 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	37.41 ms	3 - 620 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Elite Mobile	45.51 ms	3 - 846 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® X2 Elite	36.906 ms	4 - 4 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® X Elite	77.564 ms	4 - 4 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® X Elite	77.564 ms	4 - 4 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	57.015 ms	4 - 854 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	168.318 ms	1 - 659 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	77.687 ms	4 - 6 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8775P	82.94 ms	2 - 865 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8775P	82.94 ms	2 - 865 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8775P	82.94 ms	2 - 865 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® QCS9075	105.177 ms	4 - 9 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	119.423 ms	4 - 999 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA7255P	168.318 ms	1 - 659 MB	NPU
RF-DETR	QNN_DLC	float	Qualcomm® SA8295P	108.245 ms	0 - 871 MB	NPU
RF-DETR	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	45.51 ms	3 - 846 MB	NPU
RF-DETR	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	44.216 ms	5 - 599 MB	NPU
RF-DETR	TFLITE	float	Snapdragon® 8 Elite Mobile	50.68 ms	6 - 611 MB	NPU
RF-DETR	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	68.807 ms	6 - 778 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® QCS8275 (Proxy)	192.334 ms	6 - 795 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® QCS8550 (Proxy)	95.896 ms	6 - 11 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® SA8775P	100.758 ms	7 - 800 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® SA8775P	100.758 ms	7 - 800 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® SA8775P	100.758 ms	7 - 800 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® QCS9075	116.925 ms	6 - 77 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® QCS8450 (Proxy)	136.182 ms	0 - 663 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® SA7255P	192.334 ms	6 - 795 MB	NPU
RF-DETR	TFLITE	float	Qualcomm® SA8295P	120.485 ms	7 - 837 MB	NPU
RF-DETR	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	50.68 ms	6 - 611 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.