A Survey on Deep Learning Technique for Video Segmentation

Image segmentation is a significant issue in computer vision and image processing, and many segmentation techniques are found in the literature. It has several applications, including robotic perception, augmented reality, medical image analysis, scene interpretation, photo compression and video surveillance. With Deep Learning (DL) becoming increasingly popular, new techniques for segmenting images with DL models have been created. The research discusses a variety of novel approaches in instance and semantic segmentation in this recently published literature review, which includes pyramid and multi-scale-based methods, convolutional pixel-labeling networks, generative and visual attention models and encoder-decoder architectures in competitive settings. This work examines some of the most well-known data sets, compares several DL-based segmentation models' advantages, disadvantages, and connections, assesses the outcomes, and discusses potential future research directions.