The paper proposes SLiMe, a method that leverages large vision-language models to segment images with as few as one annotated sample. SLiMe outperforms existing one-shot and few-shot segmentation methods.

https://arxiv.org/abs//2309.03179

YouTube: https://www.youtube.com/@ArxivPapers

PODCASTS:

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers