CS231n: Deep Learning for Computer Vision

It builds convolutional networks and modern vision models from the ground up, and the assignments are demanding enough that finishing them genuinely teaches you the craft. Fei-Fei Li and Andrej Karpathy's course is where CNNs for image recognition went mainstream, and the material still sets a benchmark few online courses match.

It is not for beginners and not a quick applied tutorial: it assumes calculus, linear algebra and prior ML, and there is no certificate from the free materials. If you are new to ML, start with a foundations course first. The public recordings are a few years old, so some newer vision architectures sit outside the core lectures, though the fundamentals still hold (as of 2026).

CS231n is a highly influential academic course in machine learning — the course where convolutional neural networks for image recognition became mainstream, first delivered by Fei-Fei Li and Andrej Karpathy at Stanford. It covers convolutional network architecture in depth, backpropagation, training techniques (batch normalization, dropout, learning rate scheduling), object detection frameworks (YOLO, Faster R-CNN), image segmentation, recurrent networks for sequence tasks, and generative models (GANs, VAEs).