Search⌘ K
AI Features
Home
Courses
Mastering Self-Supervised Algorithms for Learning without Labels

Solving Jigsaw Puzzles

Explore how to use jigsaw puzzles as a pretext task in self-supervised learning to help neural networks develop visuospatial understanding. Learn to generate shuffled image patches, apply transformations, and train a context-free network to predict permutation order, enhancing representation learning without labeled data.

We'll cover the following...

Jigsaw puzzles

Similar to predicting the relative position of patches, this pretext task involves asking a neural network to solve jigsaw puzzles to develop a visuospatial representation of objects in the image. As shown in the figure below, the input image XiX_i is first split into a 3×33\times3 grid, and all nine patches, Xi=[Xip1,Xip2,...Xip9]X_i = [X_i^{p_1}, X_i^{p_2}, ... X_i^{p_9}] ...