ResNet-Mixed-Convolution: Optimized for Qualcomm Devices

ResNet Mixed Convolutions is a network with a mixture of 2D and 3D convolutions used for video understanding.

This is based on the implementation of ResNet-Mixed-Convolution 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	w8a16	Universal	QAIRT 2.45, ONNX Runtime 1.25.0	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
QNN_DLC	w8a16	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

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

Model Details

Model Type: Model_use_case.video_classification

Model Stats:

• Model checkpoint: Kinetics-400
• Input resolution: 112x112
• Number of parameters: 11.7M
• Model size (float): 44.6 MB
• Model size (w8a16): 11.5 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® X2 Elite	135.845 ms	201 - 201 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® X Elite	70.28 ms	170 - 170 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Gen 3 Mobile	50.341 ms	12 - 711 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Gen 1 Mobile	131.584 ms	0 - 498 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS8550 (Proxy)	69.642 ms	12 - 16 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS8450	131.584 ms	0 - 498 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Elite Mobile	42.47 ms	2 - 543 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	30.63 ms	4 - 564 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS9075	129.686 ms	11 - 57 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS8750	42.47 ms	2 - 543 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS7181	70.28 ms	170 - 170 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® X2 Elite	36.529 ms	207 - 207 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® X Elite	44.955 ms	176 - 176 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	33.11 ms	6 - 576 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Gen 1 Mobile	82.24 ms	7 - 574 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	43.691 ms	0 - 26 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS8450	82.24 ms	7 - 574 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS9075	46.276 ms	6 - 51 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	19.329 ms	0 - 407 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Elite Mobile	26.686 ms	2 - 382 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS8750	26.686 ms	2 - 382 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS7181	44.955 ms	176 - 176 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® X2 Elite	37.985 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® X Elite	70.836 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	50.769 ms	12 - 716 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	133.119 ms	3 - 513 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8275	494.37 ms	1 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	69.307 ms	12 - 15 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8450	133.119 ms	3 - 513 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Elite Mobile	41.797 ms	12 - 530 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA8295P	137.666 ms	0 - 355 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	30.542 ms	11 - 569 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA7255P	494.37 ms	1 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS9075	131.044 ms	12 - 25 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8750	41.797 ms	12 - 530 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS7181	70.836 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® X2 Elite	38.073 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® X Elite	52.111 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	38.826 ms	6 - 722 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Gen 1 Mobile	76.558 ms	5 - 723 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS6490	180.805 ms	8 - 15 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8275	159.745 ms	6 - 468 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	50.627 ms	6 - 8 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8450	76.558 ms	5 - 723 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS9075	47.492 ms	1 - 8 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	21.284 ms	6 - 437 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Elite Mobile	30.002 ms	6 - 399 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA8295P	77.599 ms	0 - 407 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	89.291 ms	5 - 459 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA7255P	159.745 ms	6 - 468 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCM6690	1151.142 ms	6 - 513 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS7790	89.291 ms	5 - 459 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8750	30.002 ms	6 - 399 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS7181	52.111 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1222.408 ms	0 - 837 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	1789.43 ms	1 - 654 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8275	2982.123 ms	0 - 651 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1552.562 ms	0 - 4 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8775P	187519.464 ms	863 - 879 MB	CPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8650P	187519.464 ms	863 - 879 MB	CPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8255P	187519.464 ms	863 - 879 MB	CPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8450	1789.43 ms	1 - 654 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Elite Mobile	1070.064 ms	0 - 643 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8295P	1816.448 ms	1 - 504 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	965.706 ms	0 - 657 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA7255P	2982.123 ms	0 - 651 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS9075	1627.872 ms	0 - 42 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8750	1070.064 ms	0 - 643 MB	NPU

License

• The license for the original implementation of ResNet-Mixed-Convolution can be found here.

References

• A Closer Look at Spatiotemporal Convolutions for Action 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.