metadata

tags:
  - model_hub_mixin
  - pytorch_model_hub_mixin
license: mit

Retinaface

Model Details

Model Description

Retinaface is a state-of-the-art face detection model built using PyTorch. It accurately detects faces in images and returns bounding boxes around detected faces. The model is designed to work efficiently on a wide range of images, including those with varying lighting conditions, occlusions, and face orientations.

License: MIT
License Link: MIT License

Model Sources

Repository: Pytorch_Retinaface
Paper: RetinaFace: Single-stage Dense Face Localisation in the Wild

Model Architecture

The Retinaface model utilizes a deep convolutional neural network architecture with multiple layers. It uses mobilenet0.25 as the backbone network (only 1.7M parameters) but can also use resnet50 as the backbone to achieve better results. It includes additional layers for feature extraction and bounding box prediction.

Intended Use

This model is intended for use in applications requiring face detection, such as:

Security systems
Augmented reality
Image processing pipelines
Photo management applications

Evaluation Results

The model was evaluated on the WIDER FACE dataset

Limitations and Biases

While the Retinaface model performs well in many conditions, it may have limitations, including:

Reduced accuracy in detecting faces with heavy occlusions Potential biases towards the demographic distribution of the training dataset Ethical Considerations When using face detection technologies, it is essential to consider the ethical implications, such as privacy concerns and potential biases. Ensure that the use of this model complies with relevant regulations and guidelines.

Citation

If you use the Retinaface model in your research or application, please cite the following paper:

@misc{deng2019retinafacesinglestagedenseface,
      title={RetinaFace: Single-stage Dense Face Localisation in the Wild}, 
      author={Jiankang Deng and Jia Guo and Yuxiang Zhou and Jinke Yu and Irene Kotsia and Stefanos Zafeiriou},
      year={2019},
      eprint={1905.00641},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/1905.00641}
}

Acknowledgements

We thank the contributors and the open-source community for their valuable support in developing this model. Special thanks to the authors of the original Retinaface paper and the WIDER FACE dataset.