ResNeXt50: Optimized for Qualcomm Devices

ResNeXt50 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 ResNeXt50 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
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	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 ResNeXt50 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 ResNeXt50 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: 25.0M
• Model size (float): 95.4 MB
• Model size (w8a8): 24.8 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNeXt50	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.132 ms	1 - 85 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® X2 Elite	1.098 ms	50 - 50 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® X Elite	2.422 ms	50 - 50 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.669 ms	0 - 150 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.284 ms	1 - 9 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS9075	3.453 ms	0 - 4 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.383 ms	0 - 85 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.603 ms	0 - 79 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X2 Elite	0.505 ms	25 - 25 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X Elite	1.281 ms	25 - 25 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.814 ms	0 - 100 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS6490	46.441 ms	8 - 24 MB	CPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.128 ms	0 - 33 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS9075	1.233 ms	0 - 3 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCM6690	26.538 ms	3 - 12 MB	CPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.692 ms	0 - 68 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	20.706 ms	5 - 14 MB	CPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.157 ms	1 - 75 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X2 Elite	1.447 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X Elite	2.697 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.811 ms	1 - 136 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	12.03 ms	1 - 71 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	2.573 ms	1 - 2 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA8775P	3.825 ms	1 - 73 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS9075	3.636 ms	3 - 5 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	4.022 ms	0 - 113 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA7255P	12.03 ms	1 - 71 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA8295P	4.111 ms	0 - 56 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.472 ms	0 - 73 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.538 ms	0 - 75 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.613 ms	0 - 0 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X Elite	1.233 ms	0 - 0 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.78 ms	0 - 99 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS6490	3.103 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	2.454 ms	0 - 71 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.11 ms	0 - 129 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA8775P	1.444 ms	0 - 72 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS9075	1.199 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCM6690	7.622 ms	0 - 194 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	1.477 ms	0 - 100 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA7255P	2.454 ms	0 - 71 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA8295P	1.762 ms	0 - 68 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.635 ms	0 - 68 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.373 ms	0 - 75 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.178 ms	0 - 116 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.771 ms	0 - 177 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8275 (Proxy)	12.032 ms	0 - 112 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	2.473 ms	0 - 2 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8775P	3.84 ms	0 - 113 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS9075	3.686 ms	0 - 52 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8450 (Proxy)	4.004 ms	0 - 151 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA7255P	12.032 ms	0 - 112 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8295P	4.146 ms	0 - 87 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.435 ms	0 - 116 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.485 ms	0 - 73 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.676 ms	0 - 99 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS6490	2.807 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	2.215 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.894 ms	0 - 2 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8775P	1.276 ms	0 - 71 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS9075	0.992 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCM6690	7.047 ms	0 - 198 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	1.286 ms	0 - 99 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA7255P	2.215 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8295P	1.556 ms	0 - 65 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.551 ms	0 - 64 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.218 ms	0 - 70 MB	NPU

License

• The license for the original implementation of ResNeXt50 can be found here.

References

• Aggregated Residual Transformations for Deep Neural Networks
• 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.