DETR-ResNet101: 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 DETR-ResNet101 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 1.25.0	Download
ONNX	w8a16_mixed_int16	Universal	QAIRT 2.45, 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 DETR-ResNet101 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 DETR-ResNet101 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Model checkpoint: ResNet101
• Input resolution: 480x480
• Number of parameters: 60.3M
• Model size (float): 230 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DETR-ResNet101	ONNX	float	Snapdragon® X2 Elite	12.68 ms	176 - 176 MB	NPU
DETR-ResNet101	ONNX	float	Snapdragon® X Elite	24.833 ms	145 - 145 MB	NPU
DETR-ResNet101	ONNX	float	Snapdragon® 8 Gen 3 Mobile	18.35 ms	5 - 462 MB	NPU
DETR-ResNet101	ONNX	float	Snapdragon® 8 Gen 1 Mobile	49.303 ms	5 - 359 MB	NPU
DETR-ResNet101	ONNX	float	Qualcomm® QCS8550 (Proxy)	24.54 ms	0 - 126 MB	NPU
DETR-ResNet101	ONNX	float	Qualcomm® QCS8450	49.303 ms	5 - 359 MB	NPU
DETR-ResNet101	ONNX	float	Snapdragon® 8 Elite Mobile	14.615 ms	3 - 327 MB	NPU
DETR-ResNet101	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	11.692 ms	3 - 334 MB	NPU
DETR-ResNet101	ONNX	float	Qualcomm® QCS9075	41.974 ms	5 - 50 MB	NPU
DETR-ResNet101	ONNX	float	Qualcomm® QCS8750	14.615 ms	3 - 327 MB	NPU
DETR-ResNet101	ONNX	float	Qualcomm® QCS7181	24.833 ms	145 - 145 MB	NPU
DETR-ResNet101	ONNX	w8a16_mixed_int16	Snapdragon® 8 Gen 3 Mobile	57.257 ms	17 - 594 MB	NPU
DETR-ResNet101	ONNX	w8a16_mixed_int16	Snapdragon® 8 Elite Mobile	47.629 ms	13 - 449 MB	NPU
DETR-ResNet101	ONNX	w8a16_mixed_int16	Snapdragon® 8 Elite Gen 5 Mobile	39.118 ms	13 - 499 MB	NPU
DETR-ResNet101	ONNX	w8a16_mixed_int16	Qualcomm® QCS8750	47.629 ms	13 - 449 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® X2 Elite	12.755 ms	5 - 5 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® X Elite	27.1 ms	5 - 5 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	19.681 ms	0 - 437 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	54.832 ms	4 - 342 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS8275	138.077 ms	2 - 319 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	26.631 ms	5 - 7 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS8450	54.832 ms	4 - 342 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® 8 Elite Mobile	15.101 ms	5 - 327 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® SA8295P	43.218 ms	0 - 233 MB	NPU
DETR-ResNet101	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	10.976 ms	5 - 347 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® SA7255P	138.077 ms	2 - 319 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS9075	46.088 ms	5 - 11 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS8750	15.101 ms	5 - 327 MB	NPU
DETR-ResNet101	QNN_DLC	float	Qualcomm® QCS7181	27.1 ms	5 - 5 MB	NPU
DETR-ResNet101	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	19.785 ms	0 - 492 MB	NPU
DETR-ResNet101	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	54.48 ms	0 - 390 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® QCS8275	138.318 ms	0 - 368 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® QCS8550 (Proxy)	26.912 ms	0 - 3 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® SA8775P	1042.992 ms	0 - 12 MB	CPU
DETR-ResNet101	TFLITE	float	Qualcomm® SA8650P	1042.992 ms	0 - 12 MB	CPU
DETR-ResNet101	TFLITE	float	Qualcomm® SA8255P	1042.992 ms	0 - 12 MB	CPU
DETR-ResNet101	TFLITE	float	Qualcomm® QCS8450	54.48 ms	0 - 390 MB	NPU
DETR-ResNet101	TFLITE	float	Snapdragon® 8 Elite Mobile	15.356 ms	0 - 372 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® SA8295P	43.342 ms	0 - 274 MB	NPU
DETR-ResNet101	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	11.014 ms	0 - 393 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® SA7255P	138.318 ms	0 - 368 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® QCS9075	41.854 ms	0 - 125 MB	NPU
DETR-ResNet101	TFLITE	float	Qualcomm® QCS8750	15.356 ms	0 - 372 MB	NPU

License

• The license for the original implementation of DETR-ResNet101 can be found here.

References

• End-to-End Object Detection with Transformers
• 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.