Occlusion-aware FERNet: an optimized patch-based adaptive residual network with attention mechanism for occlusion-aware facial expression recognition

The traditional FER techniques have provided higher recognition accuracy during FER, but the utilization of memory storage size of the model is high, which may degrade the performance of the FER. In order to address these challenges, an adaptive occlusion-aware FER technique is introduced. The occlusion is significantly detected by employing convolutional neural network (CNN). The face inpainting is accomplished via generative adversarial network (GAN). Lastly, the facial expressions are determined using the patch-based adaptive residual network with attention mechanism (PAResAM), where some hyperparameters are tuned optimally by using an improved arithmetic optimization algorithm (IAOA). The performance is evaluated and measured with diverse metrics. Thus, the outcome of the model ensures that it exploits a higher detection accuracy value.