Stanford CS231n: Convolutional Neural Networks for Visual Recognition

The definitive deep learning course for computer vision, created by Fei-Fei Li and Andrej Karpathy.

About This Course

Stanford CS231n is one of the most influential deep learning courses ever offered. Originally developed by Fei-Fei Li and taught by Andrej Karpathy (later Head of AI at Tesla), it covers convolutional neural networks in depth and their application to visual recognition tasks.

The course provides a thorough grounding in both the theory and practice of deep learning for vision, starting from the fundamentals and building up to advanced architectures.

What You Will Learn

  • Image Classification: The core task, k-nearest neighbors, linear classifiers
  • Loss Functions: SVM loss, softmax/cross-entropy, regularization
  • Optimization: SGD, momentum, Adam, learning rate schedules
  • Backpropagation: Computational graphs, chain rule, gradient flow
  • Convolutional Networks: Convolution, pooling, architectures (AlexNet, VGG, ResNet, etc.)
  • Training Neural Networks: Batch normalization, dropout, data augmentation, transfer learning
  • Detection and Segmentation: R-CNN, YOLO, semantic/instance segmentation
  • Generative Models: GANs, VAEs, autoregressive models
  • Recurrent Networks: RNNs, LSTMs for sequential visual data

Prerequisites

Python and NumPy proficiency. Linear algebra, calculus, and basic probability/statistics.

Course content belongs to Stanford University.

Up next

Instructor Spotlights

Meet the legendary CS educators whose courses changed how the world learns.