Topological Analysis of Convolutional Neural Network Layers for Image Analysis

Explore the intersection of Topological Data Analysis and Convolutional Neural Networks in image analysis through this 29-minute lecture. Delve into the effectiveness of Persistent Homology in detecting subtle changes in image texture primitives caused by tampering or abnormalities. Examine various topologically sensitive texture features and investigate the impact of CNN layers on these features. Learn about traditional Machine Learning, texture-based image feature landmarks, and entropy of convolved ultrasound images. Analyze the effects of convolution layers on classification accuracy through a case study, gaining insights into the black box nature of CNN decision-making processes.