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 2.45, ONNX Runtime 1.25.0	Download
ONNX	w8a8	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	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® X2 Elite	1.112 ms	180 - 180 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® X Elite	2.215 ms	148 - 148 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Gen 3 Mobile	1.648 ms	0 - 127 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Gen 1 Mobile	3.487 ms	1 - 104 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS8550 (Proxy)	2.32 ms	0 - 59 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS8450	3.487 ms	1 - 104 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Elite Mobile	1.395 ms	0 - 73 MB	NPU
ResNeXt50	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	1.137 ms	1 - 71 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS9075	3.443 ms	1 - 46 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS8750	1.395 ms	0 - 73 MB	NPU
ResNeXt50	ONNX	float	Qualcomm® QCS7181	2.215 ms	148 - 148 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X2 Elite	0.514 ms	213 - 213 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® X Elite	1.092 ms	149 - 149 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	0.816 ms	0 - 96 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Gen 1 Mobile	1.505 ms	0 - 95 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS6490	3.089 ms	0 - 45 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	1.126 ms	0 - 43 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS8450	1.505 ms	0 - 95 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.616 ms	0 - 75 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	1.37 ms	0 - 80 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCM6690	7.817 ms	0 - 195 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS9075	1.244 ms	0 - 45 MB	NPU
ResNeXt50	ONNX	w8a8	Snapdragon® 8 Elite Mobile	0.691 ms	0 - 75 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS7790	1.37 ms	0 - 80 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS8750	0.691 ms	0 - 75 MB	NPU
ResNeXt50	ONNX	w8a8	Qualcomm® QCS7181	1.092 ms	149 - 149 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X2 Elite	1.437 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® X Elite	2.682 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	1.813 ms	0 - 134 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	4.017 ms	0 - 114 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8275	12.044 ms	1 - 72 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	2.561 ms	0 - 8 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8450	4.017 ms	0 - 114 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite Mobile	1.463 ms	1 - 72 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA8295P	4.107 ms	1 - 57 MB	NPU
ResNeXt50	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	1.173 ms	1 - 74 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® SA7255P	12.044 ms	1 - 72 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS9075	3.65 ms	1 - 3 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS8750	1.463 ms	1 - 72 MB	NPU
ResNeXt50	QNN_DLC	float	Qualcomm® QCS7181	2.682 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X2 Elite	0.623 ms	0 - 0 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® X Elite	1.226 ms	1 - 1 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	0.791 ms	0 - 101 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Gen 1 Mobile	1.461 ms	0 - 103 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS6490	3.111 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8275	2.462 ms	0 - 72 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	1.085 ms	0 - 2 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8450	1.461 ms	0 - 103 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.543 ms	0 - 76 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	1.367 ms	0 - 77 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCM6690	7.753 ms	0 - 196 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS9075	1.21 ms	2 - 4 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA7255P	2.462 ms	0 - 72 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	0.645 ms	0 - 74 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® SA8295P	1.756 ms	0 - 71 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS7790	1.367 ms	0 - 77 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS8750	0.645 ms	0 - 74 MB	NPU
ResNeXt50	QNN_DLC	w8a8	Qualcomm® QCS7181	1.226 ms	1 - 1 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1.76 ms	0 - 175 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	4.016 ms	0 - 152 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8275	11.99 ms	0 - 112 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8550 (Proxy)	2.519 ms	0 - 3 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8775P	18.298 ms	0 - 20 MB	GPU
ResNeXt50	TFLITE	float	Qualcomm® SA8650P	18.298 ms	0 - 20 MB	GPU
ResNeXt50	TFLITE	float	Qualcomm® SA8255P	18.298 ms	0 - 20 MB	GPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8450	4.016 ms	0 - 152 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite Mobile	1.426 ms	0 - 114 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA8295P	4.164 ms	0 - 87 MB	NPU
ResNeXt50	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1.189 ms	0 - 116 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® SA7255P	11.99 ms	0 - 112 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS9075	3.693 ms	0 - 52 MB	NPU
ResNeXt50	TFLITE	float	Qualcomm® QCS8750	1.426 ms	0 - 114 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	0.668 ms	0 - 99 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Gen 1 Mobile	1.276 ms	0 - 97 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS6490	2.838 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8275	2.253 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	0.894 ms	0 - 2 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8775P	18.851 ms	1 - 25 MB	GPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8650P	18.851 ms	1 - 25 MB	GPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8255P	18.851 ms	1 - 25 MB	GPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8450	1.276 ms	0 - 97 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	0.496 ms	0 - 68 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1.215 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCM6690	7.34 ms	0 - 198 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS9075	0.995 ms	0 - 27 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA7255P	2.253 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	0.552 ms	0 - 67 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® SA8295P	1.571 ms	0 - 64 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS7790	1.215 ms	0 - 70 MB	NPU
ResNeXt50	TFLITE	w8a8	Qualcomm® QCS8750	0.552 ms	0 - 67 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.