ResNet-2Plus1D: Optimized for Qualcomm Devices

ResNet (2+1)D Convolutions is a network which explicitly factorizes 3D convolution into two separate and successive operations, a 2D spatial convolution and a 1D temporal convolution. It used for video understanding applications.

This is based on the implementation of ResNet-2Plus1D 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	w8a8	Universal	QAIRT 2.42, 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 ResNet-2Plus1D 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-2Plus1D 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: 31.5M
• Model size (float): 120 MB
• Model size (w8a8): 30.8 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	33.515 ms	12 - 538 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X2 Elite	34.135 ms	169 - 169 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X Elite	66.161 ms	169 - 169 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	52.951 ms	12 - 787 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS8550 (Proxy)	68.26 ms	12 - 131 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	41.863 ms	2 - 514 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS9075	110.494 ms	11 - 68 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS8750	41.863 ms	2 - 514 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS7181	66.161 ms	169 - 169 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	11.874 ms	3 - 378 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X2 Elite	12.181 ms	209 - 209 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X Elite	25.038 ms	179 - 179 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	19.592 ms	3 - 481 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS6490	1485.723 ms	330 - 358 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	25.113 ms	0 - 53 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	1397.974 ms	180 - 187 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCM6690	1469.052 ms	317 - 325 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	15.91 ms	0 - 368 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS9075	23.251 ms	3 - 48 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS7790	1397.974 ms	180 - 187 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS8750	15.91 ms	0 - 368 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS7181	25.038 ms	179 - 179 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	29.373 ms	0 - 493 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X2 Elite	34.398 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X Elite	66.836 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	51.98 ms	8 - 715 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8275	424.819 ms	1 - 493 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	69.925 ms	12 - 817 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8775P	111.024 ms	1 - 490 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8650P	111.024 ms	1 - 490 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8255P	111.024 ms	1 - 490 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	153.146 ms	3 - 595 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8295P	118.455 ms	0 - 398 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	41.774 ms	12 - 499 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA7255P	424.819 ms	1 - 493 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS9075	111.854 ms	12 - 25 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8750	41.774 ms	12 - 499 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS7181	66.836 ms	12 - 12 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	13.12 ms	3 - 383 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X2 Elite	13.433 ms	3 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X Elite	26.994 ms	3 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	21.099 ms	3 - 466 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS6490	87.688 ms	3 - 7 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8275	75.301 ms	2 - 372 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	27.766 ms	3 - 6 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8775P	24.455 ms	3 - 371 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8650P	24.455 ms	3 - 371 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8255P	24.455 ms	3 - 371 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA7255P	75.301 ms	2 - 372 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	46.361 ms	3 - 390 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8295P	41.941 ms	1 - 368 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCM6690	423.138 ms	3 - 426 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	16.908 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS9075	24.213 ms	0 - 5 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	47.71 ms	3 - 459 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS7790	46.361 ms	3 - 390 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8750	16.908 ms	3 - 373 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS7181	26.994 ms	3 - 3 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	1320.002 ms	0 - 568 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	1481.178 ms	0 - 801 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8275	3622.62 ms	1 - 579 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	1935.795 ms	1 - 4 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8775P	1954.941 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8650P	1954.941 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8255P	1954.941 ms	1 - 580 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	2402.574 ms	2 - 741 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8295P	2446.863 ms	1 - 541 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	1400.198 ms	0 - 572 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA7255P	3622.62 ms	1 - 579 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS9075	2051.948 ms	0 - 81 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8750	1400.198 ms	0 - 572 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	3476.116 ms	0 - 1435 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	3092.161 ms	1 - 1624 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS6490	8769.594 ms	1441 - 2233 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8275	8115.715 ms	3 - 1383 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	4138.566 ms	1 - 5 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8775P	4286.515 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8650P	4286.515 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8255P	4286.515 ms	2 - 1381 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA7255P	8115.715 ms	3 - 1383 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8295P	4660.264 ms	2 - 1091 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCM6690	8018.037 ms	1299 - 1542 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2824.929 ms	1 - 1682 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS9075	2951.371 ms	0 - 74 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	4775.334 ms	0 - 1338 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8750	2824.929 ms	1 - 1682 MB	NPU

License

• The license for the original implementation of ResNet-2Plus1D 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.