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 1.24.1	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-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® X Elite	13.549 ms	63 - 63 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Gen 3 Mobile	9.62 ms	1 - 253 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS8550 (Proxy)	13.183 ms	0 - 80 MB	NPU
ResNet-3D	ONNX	float	Qualcomm® QCS9075	25.631 ms	2 - 7 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Elite For Galaxy Mobile	7.903 ms	0 - 195 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	6.042 ms	2 - 200 MB	NPU
ResNet-3D	ONNX	float	Snapdragon® X2 Elite	7.051 ms	64 - 64 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X Elite	4.778 ms	32 - 32 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Gen 3 Mobile	3.368 ms	0 - 246 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS6490	271.442 ms	56 - 133 MB	CPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS8550 (Proxy)	4.504 ms	1 - 3 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCS9075	4.709 ms	0 - 3 MB	NPU
ResNet-3D	ONNX	w8a8	Qualcomm® QCM6690	329.329 ms	54 - 61 MB	CPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.752 ms	0 - 183 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® 7 Gen 4 Mobile	282.538 ms	41 - 48 MB	CPU
ResNet-3D	ONNX	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.971 ms	0 - 174 MB	NPU
ResNet-3D	ONNX	w8a8	Snapdragon® X2 Elite	2.216 ms	33 - 33 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X Elite	14.13 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	9.789 ms	0 - 285 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8275 (Proxy)	91.627 ms	2 - 214 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8550 (Proxy)	13.362 ms	2 - 4 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8775P	24.064 ms	1 - 198 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS9075	26.87 ms	4 - 8 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® QCS8450 (Proxy)	28.089 ms	0 - 243 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA7255P	91.627 ms	2 - 214 MB	NPU
ResNet-3D	QNN_DLC	float	Qualcomm® SA8295P	26.356 ms	0 - 194 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite For Galaxy Mobile	7.832 ms	0 - 215 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	5.664 ms	2 - 209 MB	NPU
ResNet-3D	QNN_DLC	float	Snapdragon® X2 Elite	7.092 ms	2 - 2 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X Elite	4.854 ms	1 - 1 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Gen 3 Mobile	3.399 ms	0 - 123 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS6490	19.756 ms	1 - 3 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8275 (Proxy)	14.856 ms	1 - 60 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8550 (Proxy)	4.54 ms	1 - 2 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8775P	4.607 ms	1 - 60 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS9075	4.832 ms	1 - 3 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCM6690	93.456 ms	1 - 184 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® QCS8450 (Proxy)	6.271 ms	1 - 126 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA7255P	14.856 ms	1 - 60 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Qualcomm® SA8295P	8.281 ms	0 - 56 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	2.688 ms	0 - 177 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 7 Gen 4 Mobile	7.781 ms	1 - 168 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	1.898 ms	1 - 65 MB	NPU
ResNet-3D	QNN_DLC	w8a8	Snapdragon® X2 Elite	2.479 ms	1 - 1 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	205.058 ms	0 - 278 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8275 (Proxy)	539.54 ms	0 - 219 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8550 (Proxy)	311.778 ms	0 - 2 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8775P	276.899 ms	0 - 219 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS9075	298.611 ms	0 - 69 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® QCS8450 (Proxy)	299.55 ms	0 - 278 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA7255P	539.54 ms	0 - 219 MB	NPU
ResNet-3D	TFLITE	float	Qualcomm® SA8295P	336.973 ms	0 - 214 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite For Galaxy Mobile	195.525 ms	0 - 224 MB	NPU
ResNet-3D	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	186.887 ms	0 - 223 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Gen 3 Mobile	177.962 ms	0 - 322 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8275 (Proxy)	492.304 ms	0 - 252 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8550 (Proxy)	242.1 ms	0 - 3 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8775P	246.345 ms	0 - 280 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS9075	209.35 ms	0 - 67 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCM6690	1476.196 ms	656 - 812 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® QCS8450 (Proxy)	267.097 ms	0 - 359 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA7255P	492.304 ms	0 - 252 MB	NPU
ResNet-3D	TFLITE	w8a8	Qualcomm® SA8295P	256.85 ms	0 - 253 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite For Galaxy Mobile	134.964 ms	0 - 251 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 7 Gen 4 Mobile	1185.893 ms	662 - 723 MB	NPU
ResNet-3D	TFLITE	w8a8	Snapdragon® 8 Elite Gen 5 Mobile	164.735 ms	0 - 299 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.