Giorgos Tolias

gtolias.bsky.social

07 September 2026 04:38

Choices in the image acquisition and processing pipeline leave traces in the representation space of foundational vision encoders. At ECCV 2026 we discuss about the what, the why, and the how to fix it. @eccv.bsky.social #ECCV2026

Lucas Degeorge

lucasdegeorge.bsky.social

03 September 2026 12:30

🚀 New paper: Balancing Frequencies and Pixels in Flow Matching We tackle the low-frequency bias in pixel-space flow matching and train JiT up to 40% faster without any architectural changes. 📄 Read it here: arxiv.org/abs/2609.02748

Klara Janouskova

klara-cz.bsky.social

03 September 2026 12:29

🐕 🍄 🎻 🚜 ReImageNet is out. 🚜 🎻🍄� We reannotated the whole ImageNet-1k val set from scratch: 50,000 images, multilabel, 96,051 bounding boxes, new definitions for all 1,000 classes, five semantic attributes. Browse every annotation: vrg.fel.cvut.cz/reimagenet Paper: arxiv.org/abs/2608.13783

Mert Bulent Sariyildiz

mbsariyildiz.bsky.social

02 September 2026 04:23

🧐 Can you distill knowledge from 4 vision teachers into one student using ZERO real images? 🥳 Turns out: YES, and surprisingly well. 📣 Our ECCV 2026 paper "IDeaL" closes most of the gap with real-image distillation, using optimized structured noise.

Christian Wolf

chriswolfvision.bsky.social

01 September 2026 04:03

Babysitting my neural model through the learning process by finding the right hyper parameters and annealing schedules...