DETR-ResNet50: 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-ResNet50 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 2.50, ONNX Runtime 1.27.1	Download
ONNX	w8a16_mixed_fp16	Universal	QAIRT 2.50, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.50	Download
QNN_DLC	w8a16_mixed_fp16	Universal	QAIRT 2.50	Download
TFLITE	float	Universal	QAIRT 2.50	Download

For more device-specific assets and performance metrics, visit DETR-ResNet50 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-ResNet50 on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

• Input resolution: 480x480
• Model checkpoint: ResNet50
• Model size (float): 158 MB
• Number of parameters: 41.4M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
DETR-ResNet50	ONNX	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	8.257 ms	3 - 278 MB	NPU
DETR-ResNet50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	10.048 ms	3 - 259 MB	NPU
DETR-ResNet50	ONNX	float	Snapdragon® X2 Elite	8.215 ms	5 - 5 MB	NPU
DETR-ResNet50	ONNX	float	Snapdragon® X Elite	16.571 ms	78 - 78 MB	NPU
DETR-ResNet50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	12.998 ms	0 - 362 MB	NPU
DETR-ResNet50	ONNX	float	Snapdragon® 8 Gen 1 Mobile	26.965 ms	1 - 312 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	27.801 ms	5 - 13 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	16.508 ms	0 - 95 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® QCS8450	26.965 ms	1 - 312 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	25.071 ms	5 - 13 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	16.571 ms	78 - 78 MB	NPU
DETR-ResNet50	ONNX	float	Qualcomm® Dragonwing™ Q-8750	10.048 ms	3 - 259 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	24.766 ms	21 - 289 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Elite For Galaxy Mobile	31.162 ms	20 - 271 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	37.101 ms	19 - 364 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-8275	63.9 ms	20 - 23 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	44.458 ms	0 - 49 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Qualcomm® QCS8450	97.068 ms	7 - 466 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-9075	46.732 ms	23 - 25 MB	NPU
DETR-ResNet50	ONNX	w8a16_mixed_fp16	Qualcomm® Dragonwing™ Q-8750	31.162 ms	20 - 271 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	7.92 ms	5 - 278 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	10.313 ms	5 - 264 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® X2 Elite	9.25 ms	5 - 5 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® X Elite	19.24 ms	5 - 5 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	13.709 ms	3 - 341 MB	NPU
DETR-ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	28.189 ms	4 - 292 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	27.309 ms	5 - 12 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	18.874 ms	5 - 7 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® SA8650P	26.853 ms	1 - 300 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® SA8255P	26.853 ms	1 - 300 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® QCS8450	28.189 ms	4 - 292 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	27.22 ms	7 - 14 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	19.24 ms	5 - 5 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	10.313 ms	5 - 264 MB	NPU
DETR-ResNet50	QNN_DLC	float	Qualcomm® SA8295P	29.319 ms	0 - 218 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	10.826 ms	2 - 306 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Elite For Galaxy Mobile	12.726 ms	0 - 286 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Snapdragon® X2 Elite	11.199 ms	2 - 2 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Snapdragon® X Elite	23.917 ms	2 - 2 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Snapdragon® 8 Gen 3 Mobile	17.728 ms	2 - 370 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-8275	28.567 ms	2 - 8 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ QCS8550 (Proxy)	24.229 ms	2 - 5 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-9075	28.548 ms	2 - 6 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ IQ-X7181	23.917 ms	2 - 2 MB	NPU
DETR-ResNet50	QNN_DLC	w8a16_mixed_fp16	Qualcomm® Dragonwing™ Q-8750	12.726 ms	0 - 286 MB	NPU
DETR-ResNet50	TFLITE	float	Snapdragon® 8 Elite Gen 5 For Galaxy Mobile	7.948 ms	0 - 301 MB	NPU
DETR-ResNet50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	10.35 ms	0 - 290 MB	NPU
DETR-ResNet50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	13.575 ms	0 - 362 MB	NPU
DETR-ResNet50	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	27.991 ms	0 - 313 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	27.444 ms	0 - 89 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	19.107 ms	0 - 2 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® SA8650P	26.917 ms	0 - 323 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® SA8255P	26.917 ms	0 - 323 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® QCS8450	27.991 ms	0 - 313 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	26.677 ms	0 - 89 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	10.35 ms	0 - 290 MB	NPU
DETR-ResNet50	TFLITE	float	Qualcomm® SA8295P	29.311 ms	0 - 237 MB	NPU

License

• The license for the original implementation of DETR-ResNet50 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.