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 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.19.1	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.19.1	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	1.134 ms	1 - 86 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® X2 Elite	1.098 ms	50 - 50 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® X Elite	2.428 ms	50 - 50 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.663 ms	0 - 146 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.244 ms	1 - 4 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS9075	3.446 ms	0 - 4 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	1.375 ms	0 - 84 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.596 ms	0 - 79 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X2 Elite	1.177 ms	25 - 25 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X Elite	1.266 ms	25 - 25 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.807 ms	0 - 102 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS6490	46.813 ms	5 - 26 MB	CPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.1 ms	0 - 160 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS9075	1.257 ms	0 - 3 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCM6690	26.276 ms	4 - 12 MB	CPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.693 ms	0 - 68 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	20.687 ms	5 - 14 MB	CPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.185 ms	1 - 77 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X2 Elite	1.419 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X Elite	2.68 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.789 ms	0 - 139 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	11.901 ms	1 - 75 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	2.529 ms	1 - 58 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA8775P	3.776 ms	1 - 76 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS9075	3.672 ms	1 - 3 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	4.084 ms	0 - 116 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA7255P	11.901 ms	1 - 75 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA8295P	4.116 ms	0 - 53 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	1.487 ms	0 - 76 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.531 ms	0 - 73 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.683 ms	0 - 0 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X Elite	1.242 ms	0 - 0 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.807 ms	0 - 96 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS6490	3.07 ms	2 - 4 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	2.488 ms	0 - 69 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.094 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA8775P	1.494 ms	0 - 70 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS9075	1.232 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCM6690	7.617 ms	0 - 193 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	1.461 ms	0 - 98 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA7255P	2.488 ms	0 - 69 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA8295P	1.777 ms	0 - 66 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.634 ms	0 - 72 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.388 ms	0 - 73 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.195 ms	0 - 126 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.746 ms	0 - 186 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8275 (Proxy)	11.897 ms	0 - 123 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	2.466 ms	0 - 2 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8775P	3.844 ms	0 - 124 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS9075	3.729 ms	0 - 52 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8450 (Proxy)	3.979 ms	0 - 161 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA7255P	11.897 ms	0 - 123 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8295P	4.175 ms	0 - 104 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1.427 ms	0 - 124 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.492 ms	0 - 72 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.688 ms	0 - 101 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS6490	2.867 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	2.252 ms	0 - 69 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.916 ms	0 - 60 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8775P	1.297 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS9075	1.021 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCM6690	7.326 ms	0 - 198 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	1.315 ms	0 - 101 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA7255P	2.252 ms	0 - 69 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8295P	1.574 ms	0 - 64 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	0.557 ms	0 - 63 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.232 ms	0 - 69 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.