Course book
Applied Deep Learning
Four illustrated chapters, one for each video course. They explain the ideas in original words and work the key calculations by hand.
- Chapter 1Deep Learning from the Top DownFine-tuning pre-trained models, p-values worked with coin flips, the DataBlock API and deployment, SGD from scratch, data ethics, and collaborative filtering by hand.
- Chapter 2Image Processing with OpenCVImages as arrays and BGR, greyscale and resizing worked by hand, transformations and contours, colour spaces, blurs compared, masks, histograms, thresholds, edges and Haar cascades.
- Chapter 3How a Self-Driving Car SeesRoad segmentation and IoU, YOLO and non-maximum suppression, Deep SORT with a Kalman filter worked by hand, LIDAR and homogeneous transforms, multi-task, 3D and bird's-eye perception.
- Chapter 4Learning Games by Self-PlayGame interfaces, Monte Carlo tree search, the policy and value network, PUCT selection and the AlphaZero loss worked by hand, temperature, noise, parallel self-play and evaluation.