A review on deep learning for visual understanding

Deep learning algorithms are a kind of machine learning that aims to find many layers of distributed representations. To address conventional artificial intelligence issues, a number of deep learning methods have recently been suggested. This article attempts to summarize the state-of-the-art in computer vision deep learning algorithms by emphasizing contributions and difficulties from over 210 recent research publications. It begins by providing an overview of the different classifiers and their recent developments, followed by a brief description of their applications in a variety of vision applications, including image classification, object detection, image retrieval, semantic segmentation, and human pose estimation. Finally, the article outlines future trends and difficulties in neural network based design and training.