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 1.24.1	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.1	Download
QNN_DLC	float	Universal	QAIRT 2.43	Download
QNN_DLC	w8a8	Universal	QAIRT 2.43	Download
TFLITE	float	Universal	QAIRT 2.43, TFLite 2.17.0	Download
TFLITE	w8a8	Universal	QAIRT 2.43, TFLite 2.17.0	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® X Elite	12.203 ms	60 - 60 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.023 ms	2 - 300 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS8550 (Proxy)	11.987 ms	2 - 8 MB	NPU
ResNet-2Plus1D	ONNX	float	Qualcomm® QCS9075	21.906 ms	2 - 7 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.21 ms	0 - 211 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	5.52 ms	2 - 215 MB	NPU
ResNet-2Plus1D	ONNX	float	Snapdragon® X2 Elite	6.147 ms	60 - 60 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X Elite	4.544 ms	31 - 31 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.302 ms	0 - 226 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS6490	323.43 ms	98 - 127 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	4.263 ms	0 - 34 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCS9075	4.101 ms	1 - 3 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Qualcomm® QCM6690	300.872 ms	100 - 107 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.582 ms	0 - 189 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	274.471 ms	67 - 75 MB	CPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.931 ms	0 - 191 MB	NPU
ResNet-2Plus1D	ONNX	w8a8	Snapdragon® X2 Elite	2.039 ms	31 - 31 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X Elite	13.027 ms	2 - 2 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.209 ms	0 - 301 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	81.804 ms	1 - 216 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	12.66 ms	2 - 7 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8775P	21.301 ms	0 - 215 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS9075	22.983 ms	4 - 8 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	29.086 ms	0 - 280 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA7255P	81.804 ms	1 - 216 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Qualcomm® SA8295P	22.723 ms	0 - 198 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.2 ms	0 - 217 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.713 ms	2 - 230 MB	NPU
ResNet-2Plus1D	QNN_DLC	float	Snapdragon® X2 Elite	6.602 ms	2 - 2 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X Elite	4.901 ms	1 - 1 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.363 ms	0 - 221 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS6490	19.701 ms	0 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	13.281 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.574 ms	1 - 150 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.659 ms	1 - 184 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS9075	4.799 ms	1 - 3 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCM6690	71.411 ms	1 - 196 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	7.516 ms	1 - 218 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA7255P	13.281 ms	1 - 182 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Qualcomm® SA8295P	7.78 ms	1 - 181 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.545 ms	0 - 178 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.86 ms	1 - 188 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.881 ms	1 - 183 MB	NPU
ResNet-2Plus1D	QNN_DLC	w8a8	Snapdragon® X2 Elite	2.308 ms	1 - 1 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	289.093 ms	0 - 322 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8275 (Proxy)	720.668 ms	0 - 234 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	408.203 ms	0 - 2 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8775P	386.866 ms	0 - 233 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS9075	386.646 ms	0 - 66 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	440.319 ms	0 - 313 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA7255P	720.668 ms	0 - 234 MB	NPU
ResNet-2Plus1D	TFLITE	float	Qualcomm® SA8295P	468.085 ms	0 - 229 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	254.75 ms	0 - 240 MB	NPU
ResNet-2Plus1D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	262.953 ms	0 - 244 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	582.992 ms	0 - 539 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	1516.514 ms	0 - 443 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	800.022 ms	0 - 3 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8775P	792.063 ms	0 - 441 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCM6690	1605.374 ms	309 - 477 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	859.652 ms	1 - 431 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA7255P	1516.514 ms	0 - 443 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Qualcomm® SA8295P	871.277 ms	0 - 433 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	499.187 ms	0 - 522 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1253.39 ms	332 - 424 MB	NPU
ResNet-2Plus1D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	607.046 ms	0 - 472 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.