MELEX #ECCV2026 Workshop happening now at Malmömassan C1!

First talk by Prof. Josef Kittler on "Intriguing Properties of Foundation Model Representations".

@eccv.bsky.social @bazilas.bsky.social @lanalpa.bsky.social

If you run self-supervised learning on datasets where each individual image or volume is large, you have to keep your batches fairly small. Konstantinos Almpanakis from our lab found a way to train even in batches of 1, he will present it tomorrow at #ECCV2026!

This afternoon @eccv.bsky.social #ECCV2026 you can catch me at:
13:40 Wearables AI
2:30 Effective World Models for Embodied AI
3:40 CONTEXTUS
4:30 Force-Grounded Articulated Manipulation

incl my particular 🌶️ take: Don't ask what's a world model, please ask what is an action!

#ECCV2026 workshop on Physical AI starting now in Malmömässan A
phys-ai-workshop.github.io/eccv26/

SfM DL is starting with Pierre Moulin
"Revisiting the Pareto Frontier of Structure from Motion"
#ECCV2026
@eccv.bsky.social

Our paper "A Simple and Effective way to Improve FoundationPose by Enforcing a Coherent Scene" will be presented this morning at #ECCV2026 - 3rd Workshop on Scalable 3D Scene Generation and Geometric Scene Understanding - coffee break, poster#3

Meet me at #ECCV2026

today at 3pm I'll present the paper "AI-Experiment: an augmented reality art project co-created with artificial intelligence" at the #AI4VA

Paper & Code: github.com/BernhardEgge...

One of my favorite parts is, that paper & poster can be experienced in augmented reality!

2 hours to see you at our workshop.
Keynotes by:
Christian Rupprecht ,
Pierre Moulon,
Timothy Duff

And 8 papers.

Stroget 1, Quality View Hotel, 8:30
#ECCV2026

#ECCV2026 workshop!
🌐 Spatial Intelligence: from Table-top to City-scale
The 4th Perception Test Challenge

📅 9 Sep 2026 | 09:00 – 12:30 CEST

🎙 Keynotes by Jack Gallant (UC Berkeley), Laura Leal-Taixé (TUM), and Andrew Davison (Imperial).
🔗 Schedule: perception-test-challenge.github.io

We're at #ECCV2026 with papers & talks across the conference. Come say hello and learn about our latest research!

Tchaikovsky Reimagined has arrived in Malmö!

Come and visit us to collect a limited-edition souvenir (only 100 available) and leave a message in our message book :)

📍 MalmöMässan Hall C, Malmö 🇸🇪
🗓️Sept 10–11: 10:30–18:00
🗓️Sept 12: 10:30–16:00

#ECCV2026 #ECCVArt

Closing panel at #CV4GOOD, and the room is full.

Joseph Redmon, Sara Beery, Lorenz Wendt and Yann Rebois on dual use, governance, and the research agenda.

Moderated by Federico Tombari. #ECCV2026

[1/6] 📢 We are in Malmö at #ECCV2026 presenting 5 papers! 🎉

Joseph Redmon (@allen_ai) at #CV4GOOD opened with: "Can we build the panopticon, but, like, for good?"

Their answer isn't a vibe, it's mechanisms. License barring military, defense, police, surveillance and extractive use.

Invitation-only platform. Code fully open. #ECCV2026

Sara Beery at #CV4GOOD.

Active learning gets evaluated inductively, on a held-out test set. But ecologists want the whole pool labeled.

Evaluate transductively and rare classes suddenly beat common ones — look at the pink bars. #ECCV2026

Back from the break at #CV4GOOD with Sara Beery (@MIT): "Targeting Human Verification for Robust Deployment of AI in Ecology."

Room D2. #ECCV2026

(3/3) Evidence-Backed Video QA asks video models to show their work: an answer paired with precise spatio-temporal evidence, tracked masks, not just text. We introduce ST-Evidence, plus a 160k-scale training set.

arxiv.org/abs/2607.11862

#FutureOfAI #EnterpriseAI #ECCV2026

(2/3) DreamHouse asks whether vision-language models can build the real world, not just picture it. Our new benchmark grounds physical generative reasoning in timber-frame construction.

arxiv.org/abs/2603.24866

#ECCV2026

🧵 (1/3) We're heading to #ECCV2026 in Malmö, Sweden. 🇸🇪 This week, our team will present two new papers spanning physical generative reasoning and video question answering. More below ⬇️