ManySpace
Türkçe Sign in Sign up

AI & MLRecommendedAIML-CV-011

Self-Supervised Visual Representations

Learning visual features without labels: contrastive views, masked image modelling, and self-distillation.

Difficulty
4 / 5
Importance
4 / 5
Study time
1–2 h
Kind of work
Mixed

Why it matters

The learner can compare contrastive and masked-image pretext tasks and explain how each avoids representation collapse.

The route to this topic

Route AIML-CV-011
Natural Numbers What Is a Model? Variables & Constants Tasks: Regression, Classification, and Beyond Features and Targets Solving Linear Equations Function Notation & Composition Learning Paradigms: Where the Supervision Signal Comes From Classification Framing Linear Regression I: Prediction and Residuals Embeddings Images as Tensors Complex Numbers Scores and Logits Self-Supervised Learning Loss and Fitting: Squared Error Similarity in Vector Space Basic Ratios Multiclass Scores and Softmax The Artificial Neuron Objective Functions and Empirical Risk Polynomial Functions Activation Functions Likelihood Convolution Definition of the Derivative Differentiation Rules The Multilayer Perceptron Cross-Entropy Loss Filters and Feature Maps Chain Rule Forward Propagation Contrastive Learning Pooling Computational Graphs CNN Hierarchy The Chain Rule in Neural Nets Backpropagation Gradient Flow and Depth Residual Connections CNN Architectures and Residual Networks Self-Supervised Visual Representations