Convolutional Neural Networks

Explore convolutional neural networks for computer vision in this comprehensive lecture from MIT's Introduction to Deep Learning course. Delve into the rise and impact of computer vision, including applications in self-driving cars and healthcare. Learn about image representation, manual feature extraction, and the transition to learning feature representations directly from data. Understand the architecture of convolutional neural networks, including convolutional layers, non-linearity, and pooling. Discover how these networks can be applied to various tasks such as classification, semantic segmentation, and continuous control for autonomous navigation. Gain insights into the power of deep learning for computer vision and its potential to revolutionize multiple industries.

Intro
To discover from images what is present in the world, where things are, what actions are taking place, to predict and anticipate events in the world
The rise and impact of computer vision
Impact: Self-Driving Cars
Impact: Medicine, Biology, Healthcare
Images are Numbers
Tasks in Computer Vision
Manual Feature Extraction
Learning Feature Representations Can we learn a hierarchy of features directly from the data instead of hand engineering
Fully Connected Neural Network
Using Spatial Structure
Feature Extraction with Convolution
Filters to Detect X Features
The Convolution Operation
Producing Feature Maps
Convolutional Layers: Local Connectivity
Introducing Non-Linearity
Pooling
Putting it all together
An Architecture for Many Applications
Classification: Breast Cancer Screening
Semantic Segmentation: Fully Convolutional Networks
Continuous Control: Navigation from Vision
End-to-End Framework for Autonomous Navigation
Deep Learning for Computer Vision: Summary