Public experiment · Computer vision
Puzzle Cheat
Locating a puzzle piece from two photographs.
Context
The project
Puzzle Cheat is a computer-vision experiment published in 2024. From a photograph of the puzzle and another of one or more assembled pieces, the script suggests where they may fit.
Objectives
Design challenges
- 01
Turn real-world photographs into comparable inputs.
- 02
Handle rotation and scale differences before matching.
- 03
Return several candidate areas visually instead of an opaque answer.
Visual evidence
The project in view
View source fileArchitecture
Technical flow
Puzzle and piece photos → background removal → rotation and rescaling → feature extraction → visual matches
Scroll through the flow
Approach
Working steps
- 01
Remove the piece background to isolate its shape and visual content.
- 02
Test rotations and adapt the piece scale to the puzzle.
- 03
Extract and compare features to visualize possible locations.
Public repository
What the sources document
- The repository documents the full pipeline, from image preparation to result visualization.
- The project is explicitly presented as experimental and exposes its limitations.