ResNet-3D: Optimized for Qualcomm Devices

ResNet 3D is a network with 3D convolutions used for video understanding.

This is based on the implementation of ResNet-3D 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 1.24.3	Download
ONNX	w8a8	Universal	QAIRT 2.42, ONNX Runtime 1.24.3	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-3D 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-3D 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: 33.4M
• Model size (float): 127 MB
• Model size (w8a8): 32.1 MB

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
ResNet-3D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.044 ms	2 - 201 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Elite Mobile	7.906 ms	0 - 183 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X2 Elite	7.056 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X Elite	13.541 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X Elite	13.541 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.603 ms	0 - 252 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS8550 (Proxy)	13.228 ms	0 - 102 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS9075	25.586 ms	2 - 7 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.906 ms	0 - 183 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.973 ms	0 - 174 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite Mobile	2.761 ms	0 - 181 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X2 Elite	2.217 ms	33 - 33 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X Elite	4.799 ms	32 - 32 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X Elite	4.799 ms	32 - 32 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.36 ms	0 - 248 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS6490	266.814 ms	56 - 139 MB	CPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	4.533 ms	0 - 3 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS9075	4.707 ms	0 - 3 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCM6690	326.605 ms	44 - 51 MB	CPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.761 ms	0 - 181 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	286.736 ms	61 - 69 MB	CPU
ResNet-3D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	286.736 ms	61 - 69 MB	CPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.828 ms	2 - 209 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite Mobile	7.887 ms	2 - 218 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X2 Elite	7.189 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X Elite	14.131 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X Elite	14.131 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.718 ms	0 - 275 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	91.783 ms	1 - 211 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	13.368 ms	2 - 5 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8775P	24.03 ms	1 - 197 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8775P	24.03 ms	1 - 197 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8775P	24.03 ms	1 - 197 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS9075	27.638 ms	2 - 6 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	28.172 ms	0 - 238 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA7255P	91.783 ms	1 - 211 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8295P	26.392 ms	0 - 194 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.887 ms	2 - 218 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.912 ms	1 - 66 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite Mobile	2.688 ms	0 - 180 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X2 Elite	2.481 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X Elite	4.863 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X Elite	4.863 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.384 ms	0 - 122 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS6490	19.715 ms	1 - 3 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	14.928 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.545 ms	1 - 2 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.605 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.605 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.605 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS9075	4.81 ms	3 - 5 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCM6690	90.756 ms	1 - 186 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	5.795 ms	1 - 124 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA7255P	14.928 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8295P	8.304 ms	0 - 58 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.688 ms	0 - 180 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.819 ms	1 - 173 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.819 ms	1 - 173 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	187.378 ms	0 - 222 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite Mobile	193.181 ms	0 - 225 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	206.359 ms	0 - 268 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8275 (Proxy)	561.4 ms	0 - 220 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	279.58 ms	0 - 3 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8775P	297.364 ms	0 - 212 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8775P	297.364 ms	0 - 212 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8775P	297.364 ms	0 - 212 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS9075	289.01 ms	0 - 69 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	305.078 ms	1 - 261 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA7255P	561.4 ms	0 - 220 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8295P	324.2 ms	0 - 212 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	193.181 ms	0 - 225 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	157.484 ms	0 - 293 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite Mobile	134.94 ms	0 - 377 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	175.83 ms	0 - 311 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS6490	1718.786 ms	693 - 1058 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	493.053 ms	1 - 251 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	234.558 ms	0 - 3 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8775P	247.169 ms	0 - 297 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8775P	247.169 ms	0 - 297 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8775P	247.169 ms	0 - 297 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS9075	213.286 ms	0 - 67 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCM6690	1473.413 ms	659 - 814 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	254.363 ms	0 - 309 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA7255P	493.053 ms	1 - 251 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8295P	260.364 ms	0 - 251 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	134.94 ms	0 - 377 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1181.923 ms	694 - 744 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1181.923 ms	694 - 744 MB	NPU

License

• The license for the original implementation of ResNet-3D 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.