The purpose of this special topic course is to provide foundations and recent advances in designing, training, and testing of deep learning pipelines in computer vision applications using Convolutional Neural Networks (CNNs). The problem of image representation for general computer vision applications will be the core interest of this course. Students will learn how to design a deep CNN model from scratch for a particular computer vision problem, train the network with fast and high precision accuracy optimization algorithms, and optimize its hyper-parameters for fine tuning. The course syllabi will include Multi-Array (Tensor) Analysis, Convolution Layer Design, Feature Pooling, Activation Layers, Feature Normalization, Feature Classifiers, Loss-Functions, Gradient Back-Propagation, Stochastic Optimization, Generalization Problem, Data Augmentation techniques, Hyper-Parameter tuning, Data Augmentation, Transfer Learning, as well as three major applications in computer vision will be discussed in natural imaging, satellite imaging, and medical imaging.

Prerequisites:

• Python Programming (e.g. PyTorch, Tensorflow, Keras): you need to have basic knowledge/ preliminary experience with Python programming. This course, including assignments and projects, involve with Python coding and you should be feeling comfortable to further learn how to code in Python language and gain experience.
• Introduction to Calculus and Linear Algebra
• Preliminaries in Machine Learning