VICIS learns tailored image embeddings from example sets instead of generic features

serrjoa · x · 2026-07-24

Researchers introduced VICIS, a representation-learning method that uses example sets to define a tailored embedding space for what matters.

Instead of trying to preserve everything in an image, VICIS focuses on the visual signal that is actually relevant. The method is presented as useful when the thing you want to detect is easier to show than to describe, and the code and weights are available now.

Original post →

More from Research

Research channel →