The idea
I wanted to see what a scene looks like when only the things a detector recognises remain. Everyday footage becomes a collection of fragments on black, with labels becoming part of the image.
How it works
Objects are detected in photos or video, filtered by class and confidence, then kept as rectangular regions against a plain background. Labels and colours become part of the composition.
What’s inside
- Photo, folder, and video processing
- Class and confidence filtering
- Adjustable box padding and optional labels
- Black-background exports and wildlife experiments
Where it could go
A starting point for data-inspired motion treatments, machine-vision aesthetics, and batch image studies with a consistent graphic language.