Reposted by Pamela Osuna
Very cool work from former colleagues: doi.org/10.48550/arX...
They render 3D objects into 2D images, systematically varying parameters like hue, lighting, camera angle, etc.
Having tightly controlled stimulus sets like this will help compare representations across AI and biological networks.
🧪🧠
doi.org
MAPS: A Synthetic Dataset for Probing Vision Models in a Controlled 3D Scene Space
Modern vision models achieve strong performance on standard benchmarks, yet their aggregate accuracy reveals little about which scene properties drive their predictions. Existing robustness benchmarks...
