CavaFace: Optimized for Qualcomm Devices

A PyTorch-based framework for training face recognition models that generates facial embeddings for verification and identification tasks

This is based on the implementation of CavaFace 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 CavaFace 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 CavaFace on GitHub for usage instructions.

Model Details

Model Type: Model_use_case.object_detection

Model Stats:

Input resolution: 112x112
Model checkpoint: IR_SE_100_Combined_Epoch_24.pt
Model size (float): 249.96MB
Number of parameters: 65.5M

Performance Summary

Model	Runtime	Precision	Chipset	Inference Time (ms)	Peak Memory Range (MB)	Primary Compute Unit
CavaFace	ONNX	float	Snapdragon® X2 Elite	2.335 ms	1 - 1 MB	NPU
CavaFace	ONNX	float	Snapdragon® X Elite	4.303 ms	127 - 127 MB	NPU
CavaFace	ONNX	float	Snapdragon® 8 Gen 3 Mobile	3.19 ms	0 - 138 MB	NPU
CavaFace	ONNX	float	Snapdragon® 8 Gen 1 Mobile	9.533 ms	0 - 229 MB	NPU
CavaFace	ONNX	float	Qualcomm® Dragonwing™ IQ-8275	7.648 ms	0 - 4 MB	NPU
CavaFace	ONNX	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.315 ms	0 - 164 MB	NPU
CavaFace	ONNX	float	Qualcomm® QCS8450	9.533 ms	0 - 229 MB	NPU
CavaFace	ONNX	float	Qualcomm® Dragonwing™ IQ-9075	6.99 ms	0 - 3 MB	NPU
CavaFace	ONNX	float	Qualcomm® Dragonwing™ IQ-X7181	4.303 ms	127 - 127 MB	NPU
CavaFace	ONNX	float	Qualcomm® Dragonwing™ Q-8750	2.615 ms	0 - 89 MB	NPU
CavaFace	ONNX	float	Snapdragon® 8 Elite Mobile	2.615 ms	0 - 89 MB	NPU
CavaFace	ONNX	float	Snapdragon® 8 Elite Gen 5 Mobile	2.28 ms	0 - 90 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® X2 Elite	2.6 ms	0 - 0 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® X Elite	4.505 ms	1 - 1 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® 8 Gen 3 Mobile	3.189 ms	0 - 133 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® 8 Gen 1 Mobile	9.576 ms	0 - 221 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	7.594 ms	0 - 3 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-8275	24.709 ms	0 - 81 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.303 ms	0 - 2 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® SA8775P	6.92 ms	0 - 82 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® SA8650P	6.92 ms	0 - 82 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® SA8255P	6.92 ms	0 - 82 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® QCS8450	9.576 ms	0 - 221 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-9075	6.678 ms	0 - 2 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ IQ-X7181	4.505 ms	1 - 1 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® Dragonwing™ Q-8750	2.589 ms	0 - 85 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® SA7255P	24.709 ms	0 - 81 MB	NPU
CavaFace	QNN_DLC	float	Qualcomm® SA8295P	8.011 ms	0 - 198 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® 8 Elite Mobile	2.589 ms	0 - 85 MB	NPU
CavaFace	QNN_DLC	float	Snapdragon® 8 Elite Gen 5 Mobile	2.25 ms	0 - 84 MB	NPU
CavaFace	TFLITE	float	Snapdragon® 8 Gen 3 Mobile	3.136 ms	0 - 242 MB	NPU
CavaFace	TFLITE	float	Snapdragon® 8 Gen 1 Mobile	9.459 ms	0 - 331 MB	NPU
CavaFace	TFLITE	float	Qualcomm® Dragonwing™ IQ-8275	24.515 ms	0 - 96 MB	NPU
CavaFace	TFLITE	float	Qualcomm® Dragonwing™ QCS8550 (Proxy)	4.158 ms	0 - 3 MB	NPU
CavaFace	TFLITE	float	Qualcomm® SA8775P	6.898 ms	0 - 96 MB	NPU
CavaFace	TFLITE	float	Qualcomm® SA8650P	6.898 ms	0 - 96 MB	NPU
CavaFace	TFLITE	float	Qualcomm® SA8255P	6.898 ms	0 - 96 MB	NPU
CavaFace	TFLITE	float	Qualcomm® QCS8450	9.459 ms	0 - 331 MB	NPU
CavaFace	TFLITE	float	Qualcomm® Dragonwing™ IQ-9075	6.608 ms	0 - 128 MB	NPU
CavaFace	TFLITE	float	Qualcomm® Dragonwing™ Q-8750	2.582 ms	0 - 97 MB	NPU
CavaFace	TFLITE	float	Qualcomm® SA7255P	24.515 ms	0 - 96 MB	NPU
CavaFace	TFLITE	float	Qualcomm® SA8295P	7.95 ms	0 - 210 MB	NPU
CavaFace	TFLITE	float	Snapdragon® 8 Elite Mobile	2.582 ms	0 - 97 MB	NPU
CavaFace	TFLITE	float	Snapdragon® 8 Elite Gen 5 Mobile	2.23 ms	0 - 99 MB	NPU

License

The license for the original implementation of CavaFace can be found here.

References

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.

Downloads last month: -; Downloads are not tracked for this model. How to track