Mohamed Elgendy's "Deep Learning for Vision Systems" implements classic CV models using PyTorch, starting from the perceptron and gradually transitioning to CNNs, object detection (YOLO, Faster R-CNN), and image segmentation, finally touching on vision Transformers, with runnable code examples in each chapter; the mathematical derivations are slightly less detailed than those in Raschka's book, but the coverage is broad, and it discusses fully "why each architecture is designed this way."