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 2.42, ONNX Runtime 1.25.0	Download
ONNX	w8a16	Universal	QAIRT 2.42, 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® 8 Elite Gen 5 Mobile	29.755 ms	12 - 553 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® X Elite	70.315 ms	170 - 170 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Gen 3 Mobile	50.369 ms	12 - 755 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS8550 (Proxy)	70.46 ms	0 - 53 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	42.572 ms	0 - 535 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS9075	129.327 ms	11 - 57 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS8750	42.572 ms	0 - 535 MB	NPU
ResNet-Mixed-Convolution	ONNX	float	Qualcomm® QCS7181	70.315 ms	170 - 170 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	18.791 ms	6 - 391 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® X2 Elite	24.45 ms	206 - 206 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® X Elite	45.995 ms	176 - 176 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Gen 3 Mobile	33.196 ms	6 - 632 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS6490	13043.715 ms	357 - 370 MB	CPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS8550 (Proxy)	44.453 ms	0 - 44 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCM6690	6759.152 ms	252 - 259 MB	CPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 7 Gen 4 Mobile	5837.937 ms	295 - 302 MB	CPU
ResNet-Mixed-Convolution	ONNX	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	26.884 ms	0 - 377 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS9075	46.52 ms	6 - 51 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS7790	5837.937 ms	295 - 302 MB	CPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS8750	26.884 ms	0 - 377 MB	NPU
ResNet-Mixed-Convolution	ONNX	w8a16	Qualcomm® QCS7181	45.995 ms	176 - 176 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	30.9 ms	12 - 561 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® X2 Elite	38.019 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® X Elite	70.867 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	50.75 ms	12 - 721 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8275	494.363 ms	0 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	68.344 ms	12 - 14 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA8775P	126.783 ms	0 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA8650P	126.783 ms	0 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA8255P	126.783 ms	0 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	133.509 ms	2 - 510 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA7255P	494.363 ms	0 - 519 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® SA8295P	137.629 ms	0 - 355 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	41.737 ms	0 - 522 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS9075	131.064 ms	12 - 25 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS8750	41.737 ms	0 - 522 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	float	Qualcomm® QCS7181	70.867 ms	12 - 12 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Elite Gen 5 Mobile	21.369 ms	6 - 434 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® X2 Elite	26.114 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® X Elite	52.37 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Gen 3 Mobile	38.972 ms	6 - 722 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS6490	180.083 ms	3 - 10 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8275	159.784 ms	1 - 463 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8550 (Proxy)	50.65 ms	6 - 8 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA8775P	46.69 ms	6 - 409 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA8650P	46.69 ms	6 - 409 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA8255P	46.69 ms	6 - 409 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCM6690	1152.844 ms	45 - 554 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA7255P	159.784 ms	1 - 463 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 7 Gen 4 Mobile	89.119 ms	8 - 461 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® SA8295P	77.656 ms	0 - 407 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Snapdragon® 8 Elite For Galaxy Mobile	29.94 ms	6 - 399 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS9075	47.817 ms	1 - 8 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8450 (Proxy)	73.076 ms	6 - 723 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS7790	89.119 ms	8 - 461 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS8750	29.94 ms	6 - 399 MB	NPU
ResNet-Mixed-Convolution	QNN_DLC	w8a16	Qualcomm® QCS7181	52.37 ms	6 - 6 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	986.152 ms	0 - 650 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1229.982 ms	0 - 834 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8275	2973.423 ms	1 - 651 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1591.875 ms	1 - 4 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8775P	1566.284 ms	1 - 652 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8650P	1566.284 ms	1 - 652 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8255P	1566.284 ms	1 - 652 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8450 (Proxy)	1805.863 ms	0 - 658 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA7255P	2973.423 ms	1 - 651 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® SA8295P	1815.98 ms	0 - 504 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1032.833 ms	0 - 643 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS9075	1626.72 ms	0 - 42 MB	NPU
ResNet-Mixed-Convolution	TFLITE	float	Qualcomm® QCS8750	1032.833 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.