As face masks continue to be a part of our daily lives, the challenge of reconstructing occluded faces remains relevant. While several approaches have been proposed for removing masks from neutral facial images, few have explored the use of facial expressions as a dominant feature for reconstruction of expressive faces. To address this gap, we propose an expression-conditioned GAN (ECGAN) for reconstructing masked faces with a specified expression. Our approach leverages both the binary segmentation map of the mask and an expression label to generate high-quality images. To train our ECGAN in a supervised manner, we synthesize masked images using the RAFDB dataset to create non-masked-masked pairs of images for training. We evaluate of our approach on the RAFDB test set, demonstrating its effectiveness in generating realistic images that convincingly belong to the given expression class. This is further highlighted by comparing it to a baseline model and a state-of-the-art approach without expression-input. The code is available at https://github.com/SridharSola/ECGAN.

Facial expression recognition (FER) algorithms work well in constrained environments with little or no occlusion of the face. However, real-world face occlusion is prevalent, most notably with the need to use a face mask in the current Covid-19 scenario. While there are works on the problem of occlusion in FER, little has been done before on the particular face mask scenario. Moreover, the few works in this area largely use synthetically created masked FER datasets. Motivated by these challenges posed by the pandemic to FER, we present a novel dataset, the Masked Student Dataset of Expressions or MSD-E, consisting of 1,960 real-world non-masked and masked facial expression images collected from 142 individuals. Along with the issue of obfuscated facial features, we illustrate how other subtler issues in masked FER are represented in our dataset. We then provide baseline results using ResNet-18, finding that its performance dips in the non-masked case when trained for FER in the presence of masks. To tackle this, we test two training paradigms: contrastive learning and knowledge distillation, and find that they increase the model’s performance in the masked scenario while maintaining its non-masked performance. We further visualise our results using t-SNE plots and Grad-CAM, demonstrating that these paradigms capitalise on the limited features available in the masked scenario. Finally, we benchmark SOTA methods on MSD-E. The dataset is available at https://github.com/SridharSola/MSD-E.