ResNet50: Optimized for Qualcomm Devices

ResNet50 is a machine learning model that can classify images from the Imagenet dataset. It can also be used as a backbone in building more complex models for specific use cases.

This is based on the implementation of 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 2.45, ONNX Runtime 1.27.1	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a8	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download
TFLITE	w8a8	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.image_classification

Model Stats:

• Model checkpoint: Imagenet
• Input resolution: 224x224
• Number of parameters: 25.5M
• Model size (float): 97.4 MB
• Model size (w8a8): 25.1 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet50	ONNX	float	Snapdragon® X2 Elite	0.986 ms	2 - 2 MB	NPU
ResNet50	ONNX	float	Snapdragon® X Elite	1.917 ms	50 - 50 MB	NPU
ResNet50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.465 ms	0 - 79 MB	NPU
ResNet50	ONNX	float	Snapdragon® 8 Gen 1 Mobile	3.19 ms	0 - 63 MB	NPU
ResNet50	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	1.971 ms	0 - 58 MB	NPU
ResNet50	ONNX	float	Qualcomm® QCS8450	3.19 ms	0 - 63 MB	NPU
ResNet50	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	3.016 ms	0 - 4 MB	NPU
ResNet50	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	1.917 ms	50 - 50 MB	NPU
ResNet50	ONNX	float	Qualcomm® Dragonwing™ Q-8750	1.207 ms	0 - 51 MB	NPU
ResNet50	ONNX	float	Snapdragon® 8 Elite Mobile	1.207 ms	0 - 51 MB	NPU
ResNet50	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.054 ms	0 - 53 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® X2 Elite	0.41 ms	1 - 1 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® X Elite	0.833 ms	25 - 25 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.678 ms	0 - 83 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	1.204 ms	0 - 85 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® Dragonwing™ QCS6490	3.71 ms	0 - 3 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.905 ms	0 - 67 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® QCS8450	1.204 ms	0 - 85 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-9075	1.007 ms	0 - 3 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® Dragonwing™ IQ-X7181	0.833 ms	25 - 25 MB	NPU
ResNet50	ONNX	w8a8	Qualcomm® Dragonwing™ Q-8750	0.577 ms	0 - 49 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.577 ms	0 - 49 MB	NPU
ResNet50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.559 ms	0 - 50 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® X2 Elite	1.247 ms	1 - 1 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® X Elite	2.332 ms	1 - 1 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.607 ms	0 - 79 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	3.734 ms	0 - 61 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8275	10.725 ms	1 - 46 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.212 ms	0 - 3 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® SA8775P	3.35 ms	1 - 47 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® SA8650P	3.35 ms	1 - 47 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® SA8255P	3.35 ms	1 - 47 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® QCS8450	3.734 ms	0 - 61 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	3.181 ms	3 - 5 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	2.332 ms	1 - 1 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	1.268 ms	1 - 50 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® SA7255P	10.725 ms	1 - 46 MB	NPU
ResNet50	QNN_DLC	float	Qualcomm® SA8295P	3.642 ms	0 - 28 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.268 ms	1 - 50 MB	NPU
ResNet50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.1 ms	1 - 49 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.529 ms	0 - 0 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® X Elite	0.945 ms	0 - 0 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.678 ms	0 - 78 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.193 ms	0 - 79 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS6490	3.77 ms	0 - 2 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8275	2.05 ms	0 - 45 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.893 ms	0 - 2 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® SA8775P	1.089 ms	0 - 46 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® SA8650P	1.089 ms	0 - 46 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® SA8255P	1.089 ms	0 - 46 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® QCS8450	1.193 ms	0 - 79 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-9075	0.986 ms	0 - 2 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ IQ-X7181	0.945 ms	0 - 0 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-6690	6.612 ms	0 - 164 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-7790	1.224 ms	0 - 51 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® Dragonwing™ Q-8750	0.536 ms	0 - 46 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® SA7255P	2.05 ms	0 - 45 MB	NPU
ResNet50	QNN_DLC	w8a8	Qualcomm® SA8295P	1.431 ms	0 - 42 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.536 ms	0 - 46 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.494 ms	0 - 44 MB	NPU
ResNet50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.224 ms	0 - 51 MB	NPU
ResNet50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.586 ms	0 - 112 MB	NPU
ResNet50	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	3.729 ms	0 - 97 MB	NPU
ResNet50	TFLITE	float	Qualcomm® Dragonwing™ QCS8275	10.664 ms	0 - 69 MB	NPU
ResNet50	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	2.218 ms	0 - 2 MB	NPU
ResNet50	TFLITE	float	Qualcomm® SA8775P	3.343 ms	0 - 70 MB	NPU
ResNet50	TFLITE	float	Qualcomm® SA8650P	3.343 ms	0 - 70 MB	NPU
ResNet50	TFLITE	float	Qualcomm® SA8255P	3.343 ms	0 - 70 MB	NPU
ResNet50	TFLITE	float	Qualcomm® QCS8450	3.729 ms	0 - 97 MB	NPU
ResNet50	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	3.125 ms	0 - 52 MB	NPU
ResNet50	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	1.25 ms	0 - 72 MB	NPU
ResNet50	TFLITE	float	Qualcomm® SA7255P	10.664 ms	0 - 69 MB	NPU
ResNet50	TFLITE	float	Qualcomm® SA8295P	3.636 ms	0 - 50 MB	NPU
ResNet50	TFLITE	float	Snapdragon® 8 Elite Mobile	1.25 ms	0 - 72 MB	NPU
ResNet50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.09 ms	0 - 66 MB	NPU
ResNet50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.569 ms	0 - 73 MB	NPU
ResNet50	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	1.05 ms	0 - 81 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS6490	3.649 ms	0 - 27 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8275	1.755 ms	0 - 42 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ QCS8550 (Proxy)	0.758 ms	0 - 1 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® SA8775P	0.964 ms	0 - 45 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® SA8650P	0.964 ms	0 - 45 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® SA8255P	0.964 ms	0 - 45 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® QCS8450	1.05 ms	0 - 81 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ IQ-9075	0.89 ms	0 - 27 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-6690	6.191 ms	0 - 161 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-7790	1.096 ms	0 - 49 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® Dragonwing™ Q-8750	0.485 ms	0 - 40 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® SA7255P	1.755 ms	0 - 42 MB	NPU
ResNet50	TFLITE	w8a8	Qualcomm® SA8295P	1.278 ms	0 - 40 MB	NPU
ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	0.485 ms	0 - 40 MB	NPU
ResNet50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.448 ms	0 - 47 MB	NPU
ResNet50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.096 ms	0 - 49 MB	NPU

License

• The license for the original implementation of ResNet50 can be found here.

References

• Deep Residual Learning for Image Recognition
• 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.