iiinova — where your digital self keeps evolving

Reposted by Digital Signal Processing and Image Analysis (DSB)

adn 👣 @adnrv.bsky.social · 09/07/2025

Come to our poster @icmlconf.bsky.social (Thu 17, morning) to discover more about it Or check our paper: sthalles.github.io/sop/

011

Reposted by Digital Signal Processing and Image Analysis (DSB)

adn 👣 @adnrv.bsky.social · 09/07/2025

Self-Organizing Visual Prototypes (SOP) is a new, non-parametric approach to SSL that addresses these issues by representing each region in feature space using multiple real data embeddings (support embeddings), rather than a single learnable prototype.

111

Reposted by Digital Signal Processing and Image Analysis (DSB)

adn 👣 @adnrv.bsky.social · 09/07/2025

SSL, in particular, CL, relies more and more on prototypes (learnable representations that hold some group meaning) But! overclustering, enforcing equipartitions, and underrepresentation are still problems that plague the prototypes We solve them with our method: SOP

121

Reposted by Digital Signal Processing and Image Analysis (DSB)

SFI Visual Intelligence @sfi-vi.bsky.social · 14/02/2025

Did you miss out on the latest Visual Intelligence Online Seminar on equivariant self-supervision by Aiden Durrant? 👨‍🏫 🤖 🧠 📺 You can watch Durrant's talk on the Visual Intelligence YouTube channel: youtu.be/bsUGgwDYdXQ?...

youtu.be

Equivariant Self-Supervision: Exploiting Inductive Biases of Capsule Networks: Aiden Durrant

YouTube video by SFI Visual Intelligence

022