Deep Learning for Computer Vision

Catalog description

Deep learning is a branch of machine learning which learns rich data representations simultaneously in the process of neural network optimization. Deep learning has greatly advanced state-of-the-art performance in computer vision, which covers a wide range of applications in our life, e.g., search, map, self-driving cars, etc. This course provides an introduction to deep learning with a focus on computer vision algorithms. This course first covers the details of typical deep neural network architectures for computer vision, such as fully-connected networks, convolutional neural networks, recurrent neural networks, transformers, generative models, deep reinforcement learning. The optimization algorithms, the data-efficient learning techniques and the practical strategies for training and fine-tuning deep neural networks in computer vision tasks will be introduced. It will also discuss deep learning applications in traditional computer vision problems and emerging topics, e.g. image classification, object detection, image segmentation, video classification, action detection, etc., and introduce the core concepts behind those deep learning based computer vision algorithms. In this course, students are expected to implement and train the deep neural networks, and gain a detailed understanding of the cutting-edge algorithms in various computer vision problems.