Video-MAE: Optimized for Qualcomm Devices

Video MAE (Masked Auto Encoder) is a network for doing video classification that uses the ViT (Vision Transformer) backbone.

This is based on the implementation of Video-MAE 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 2.45, ONNX Runtime 1.27.1	Download
QNN_DLC	float	Universal	QAIRT 2.45	Download
TFLITE	float	Universal	QAIRT 2.45	Download

For more device-specific assets and performance metrics, visit Video-MAE 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 Video-MAE on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.video_classification

Model Stats:

• Input resolution: 224x224
• Model checkpoint: Kinectics-400
• Model size (float): 335 MB
• Number of parameters: 87.7M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
Video-MAE	ONNX	float	Snapdragon® X2 Elite	1518.863 ms	46 - 46 MB	NPU
Video-MAE	ONNX	float	Snapdragon® X Elite	2675.374 ms	191 - 191 MB	NPU
Video-MAE	ONNX	float	Snapdragon® 8 Gen 3 Mobile	2537.857 ms	1 - 6278 MB	NPU
Video-MAE	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	2930.774 ms	46 - 95 MB	NPU
Video-MAE	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	3824.221 ms	0 - 212 MB	NPU
Video-MAE	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	2756.919 ms	46 - 95 MB	NPU
Video-MAE	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	2675.374 ms	191 - 191 MB	NPU
Video-MAE	ONNX	float	Qualcomm® Dragonwing™ Q-8750	2327.454 ms	1 - 5481 MB	NPU
Video-MAE	ONNX	float	Snapdragon® 8 Elite Mobile	2327.454 ms	1 - 5481 MB	NPU
Video-MAE	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2594.863 ms	1 - 5688 MB	NPU
Video-MAE	QNN_DLC	float	Snapdragon® X2 Elite	1961.284 ms	46 - 46 MB	NPU
Video-MAE	QNN_DLC	float	Snapdragon® X Elite	3246.968 ms	46 - 46 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	4725.75 ms	46 - 94 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5151.275 ms	46 - 50 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® SA8775P	5352.202 ms	46 - 6139 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® SA8650P	5352.202 ms	46 - 6139 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® SA8255P	5352.202 ms	46 - 6139 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	7276.539 ms	48 - 96 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	3246.968 ms	46 - 46 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	3556.566 ms	26 - 6118 MB	NPU
Video-MAE	QNN_DLC	float	Qualcomm® SA8295P	3813.452 ms	35 - 5737 MB	NPU
Video-MAE	QNN_DLC	float	Snapdragon® 8 Elite Mobile	3556.566 ms	26 - 6118 MB	NPU
Video-MAE	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	4071.688 ms	3 - 6270 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	4622.577 ms	0 - 280 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	5166.465 ms	1 - 4 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® SA8775P	5322.161 ms	2 - 5973 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® SA8650P	5322.161 ms	2 - 5973 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® SA8255P	5322.161 ms	2 - 5973 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	5258.49 ms	0 - 279 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	3540.224 ms	1 - 5953 MB	NPU
Video-MAE	TFLITE	float	Qualcomm® SA8295P	3731.908 ms	2 - 5603 MB	NPU
Video-MAE	TFLITE	float	Snapdragon® 8 Elite Mobile	3540.224 ms	1 - 5953 MB	NPU
Video-MAE	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	4089.075 ms	1 - 6110 MB	NPU

License

• The license for the original implementation of Video-MAE can be found here.

References

• Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
• 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.