Rama Ramakrishnan teaches this session of MIT's 15.773 Hands-On Deep Learning, covering how convolutional neural networks process images through pooling layers and feature extraction, then turning to transfer learning and fine-tuning as shortcuts to building working models without training from scratch. He walks through a handbags-versus-shoes image classifier as a running example, showing how a pretrained CNN can be adapted to a new task with limited data. The lecture also introduces HuggingFace, demonstrating how its model hub and tools fit into a practical computer vision workflow. Aimed at students who already have some machine learning background, the session mixes conceptual explanation with concrete implementation steps, keeping the focus on what actually changes when you fine-tune a pretrained network versus training one from zero. Part of the Spring 2024 MIT OpenCourseWare series taught by Ramakrishnan.