MaskMTL: Attribute prediction in masked facial images with deep multitask learning

INTRO

Paper Link - ARXIV

We create a masked face dataset by efficiently overlaying masks of different shape, size and textures to effectively model variability generated by wearing mask. This paper presents a deep Multi-Task Learning (MTL) approach to jointly estimate various heterogeneous attributes from a single masked facial image. Experimental results on benchmark face attribute UTKFace dataset demonstrate that the proposed approach supersedes in performance to other competing techniques. The paper was selected for PReMI2021 conference.

ARCHITECTURE DIAGRAM

REQUIREMENTS

python 3.7
tensorflow
tensorflow.keras
opencv-python
matplotlib
pandas

DATASET

The dataset used in the paper can be downloded from link. Labels

USAGE/CODE FOR USE

1. code.ipynb provided has model reparation, training, evaluation and prediction by model. Refer that to train model on your dataset.
2. Use pretrained model given in PRETRAINED MODEL section below, to get predictions on masked facial images.

PREDICTION PERFORMANCE

PRETRAINED MODEL

The pretrained model can be downloaded here and used directly for prediction.

CITE

Please cite our paper if you use or refer this code:

@inproceedings{
  title={MaskMTL: Attribute prediction in masked facial images with deep multitask learning},
  author={Prerana Mukherjee, Vinay Kaushik, Ronak Gupta, Ritika Jha, Daneshwari Kankanwadi, and Brejesh Lall},
  booktitle={PREMI2021},
  year={2021}
}