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	6.043 ms	2 - 201 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X2 Elite	7.026 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X Elite	13.522 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.662 ms	1 - 254 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS8550 (Proxy)	13.243 ms	0 - 80 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS9075	25.92 ms	2 - 7 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.905 ms	0 - 182 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.975 ms	0 - 175 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X2 Elite	2.225 ms	33 - 33 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X Elite	4.785 ms	32 - 32 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.346 ms	0 - 245 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS6490	272.474 ms	56 - 138 MB	CPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	4.537 ms	1 - 4 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS9075	4.714 ms	0 - 3 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCM6690	328.859 ms	55 - 62 MB	CPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.74 ms	0 - 174 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	283.13 ms	66 - 74 MB	CPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.825 ms	2 - 209 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X2 Elite	7.178 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X Elite	14.076 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.712 ms	0 - 277 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	91.78 ms	0 - 211 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	13.371 ms	2 - 4 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8775P	23.995 ms	1 - 197 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS9075	27.617 ms	2 - 6 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	28.302 ms	2 - 240 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA7255P	91.78 ms	0 - 211 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8295P	26.403 ms	0 - 194 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.883 ms	0 - 213 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.904 ms	0 - 66 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X2 Elite	2.496 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X Elite	4.859 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.374 ms	1 - 122 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS6490	19.785 ms	1 - 3 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	14.945 ms	1 - 62 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.571 ms	0 - 4 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.605 ms	1 - 61 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS9075	4.809 ms	1 - 3 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCM6690	92.31 ms	1 - 185 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	6.107 ms	1 - 125 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA7255P	14.945 ms	1 - 62 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8295P	8.3 ms	0 - 58 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.705 ms	1 - 180 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.835 ms	1 - 174 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	187.53 ms	0 - 222 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	208.494 ms	0 - 269 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8275 (Proxy)	561.036 ms	0 - 221 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	278.138 ms	0 - 2 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8775P	299.254 ms	0 - 213 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS9075	285.288 ms	0 - 69 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	305.95 ms	0 - 260 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA7255P	561.036 ms	0 - 221 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8295P	327.473 ms	0 - 213 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	193.817 ms	0 - 219 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	159.263 ms	0 - 293 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	178.246 ms	0 - 313 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS6490	1745.069 ms	695 - 1060 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	493.699 ms	0 - 250 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	229.649 ms	0 - 3 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8775P	247.541 ms	0 - 297 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS9075	211.159 ms	0 - 67 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCM6690	1472.84 ms	657 - 812 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	253.006 ms	0 - 308 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA7255P	493.699 ms	0 - 250 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8295P	261.745 ms	0 - 251 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	134.065 ms	0 - 384 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1171.783 ms	672 - 726 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.