TRANSFER :- DEEP INDUCTIVE NETWORK FOR FACIAL EMOTION RECOGNITION

The image-based Facial Emotion Recognition (FER) aims to classify the image into basic emotions being communicated by it. FER is one of the most prominent research areas in computer vision. Most of the existing works are aimed at high-quality images which are collected in the lab environment. These images are very different from the real-life facial emotion that leads to a lack of wild labeled data. Deep learning using transfer learning has shown promising results in computer vision in solving the problem of lack of labeled data. In the recent system, there is a great focus to overcome the lack of data issue in FER. Our paper has utilized the deep residual networks with inductive learning and self-attention module to overcome this problem. We have experimented different pre training settings and datasets for the model, which are Image Net and VGG face dataset (source datasets). The selfattention block is applied for better visual perspective to the model. Our target dataset is FER-2013, a benchmark dataset in FER. Trans FER is a deep residual network based on inductive learning and attention module. Our proposed approach has achieved superior performance than the existing state of the art models in the FER application using transfer learning.