📣 ELLIS | ELLIOT Summer School on Autonomous Driving 🚗 Super happy to announce the hottest event this summer. Apply now! 📅 September 15-18, 2026 📍Barcelona, Spain 🌐 pages.cvc.uab.es/ssadriving/ @ellis.eu, ELLIS-Barcelona, CVC-Barcelona, @valeoai.bsky.social ai.bsky.social, @elliot-eu.bsky.social

Sadra is on 🔥 this week 🧿 we present FlowIt: Global Matching for Optical Flow with Confidence-Guided Refinement arxiv.org/abs/2603.28759 as someone who published her first flow paper 10 years ago, I cannot believe how good these results are:

Sadra released the code for WarpRF 🚀 a training-free uncertainty quantification framework for radiance fields based on multi-view consistency, without training or any changes to the model.

planning to visit Zurich on March 6. let me know if you would like to grab a coffee. I will try to convince robotics researchers that they still need computer vision at the Robotics, Vision, and Controls Talks at ETH Zurich 😄

Super cool app! I'm in the "Machine Learning Researchers and AI Scientists" cluster. My nearest neighbour is @fguney.bsky.social; neighbours both here as well as our countries. Hi komşu!

while preparing for my ERC interview, I worked with a Turkish designer on my slides. I loved his work so much that I recommended him to friends, who recommended him to others, and so on. today he told me he’s been hired for 4 more projects. your work really does become your reference 😇

do I have any followers* here who happen to be ELLIS program/unit directors? to become an ELLIS scholar, I need their help with the nomination 🥰 *initially I said friends but then couldn’t stop laughing at the idea of me having friends who happen to be directors 😂

Our new E2E driving method, TransFuser v6, is out on ArXiv. It outperforms all other methods on CARLA by a wide margin, 95 DS on Bench2Drive! We show that minimizing the asymmetry between data annotator and policy is key for strong IL results. Code, models, and paper: ln2697.github.io/lead/

do you also feel bad when you realize you’re not only good at non-research tasks, but actually enjoy them? like… I shouldn’t like report-writing. I should be writing papers. and yet, talking about our achievements in the first ERC reporting period feels so good!

“Unlike a camera, which stores with equal resolution each bit of visual information, sight is highly directed. It is focused on capturing relevant information to convey meaning, not fidelity. …

The 4th AI for Robotics workshop surfaced converging themes around embodied perception, task learning & evaluation methodologies - emphasising a shift to integrated, context-aware systems. Dive into our key takeaways #AI #Robotics #spatialAI 🥽 ➡️ tinyurl.com/bvxxcn5e

The Reith Lectures might be my favourite thing that comes out around this time of year. I’m really enjoying this year’s series, Moral Revolution by Rutger Bregman. www.bbc.co.uk/programmes/a...

when you find yourself saying “if you’re old enough to know Fast R-CNN stuff” in a talk 🥹🥹

Last week’s #AI4Robotics workshop was fab & fabulously full 😊 If you missed out we've good news - 📽️ All 15 talk recordings are now online 🚀! tinyurl.com/4ns5apvd Catch up on cutting-edge work in #robot perception, action & autonomy - from #SLAM & control to computer vision & large-scale learning!

last year after ECCV in Milano I joked about how expensive Italy felt and a few ppl kindly suggested I explore other parts of the country, so here I am taking the advice!

Fatih is presenting our work Mapping like a Skeptic 🔎 tomorrow at #BMVC2025! paper: arxiv.org/abs/2508.21689 code: github.com/Fatih-Erdoga...

what a nice way to end an amazing 2-day workshop in Grenoble 😊

just wanted to clarify that we always try to report results (for sure for main results, maybe not all ablations) over 3 runs* for online evaluations as mean and std. maybe it was a mistake to show barplots without mentioning this. *probably not enough but still better than reporting only one run.

Andrea Vedaldi talks about 4D reconstruction, leveraging 4D generation, but also representations based on dynamic point maps which encode points for different viewpoints AND different time steps. AI 4 Robotics workshop at @naverlabseurope.bsky.social

Fatma Guney on thinking fast and slow for autonomous driving: how to combine bigger models with higher latency with smaller models with lower latency by forecasting rich but stale features to the future. AI 4 Robotics Workshop at @naverlabseurope.bsky.social

EuroHPC wrote a piece about our love of GPUs 😊 A EuroHPC Success Story | Clear Vision for Self-Driving Cars www.eurohpc-ju.europa.eu/eurohpc-succ...

Track-On2: Enhancing Online Point Tracking with Memory By Görkay Aydemir, Weidi Xie, @fguney.bsky.social TLDR: explicit memory for point tracking, holds decoder features, similar to what MUSt3R does for reconstruction. arxiv.org/abs/2509.19115

Our Autonomous Systems workshop series is back for its 3rd edition at CVPR 2025! Remember the queue in Seattle? Please mark the workshop in your CVPR registration so we have enough space this year! opendrivelab.com/cvpr2025/wor...

Our Workshop on Uncertainty Quantification for Computer Vision goes to @cvprconference.bsky.social this year! We have a super line-up of speakers and a call for papers. This is a chance for your paper to shine at #CVPR2025 ⏲️ Submission deadline: 14 March 💻 Page: uncertainty-cv.github.io/2025/

feel free to nominate yourself or anyone else at KUIS AI talks: forms.gle/nbqhKsWbC9Swoqzi8

agreed. one of the messages from his talk is studying statistics, computer science, and economics together. I find his perspective on AI very realistic. also, how he changed the direction of discussion on bias during the panel in the end, should be studied in textbooks.

in a privileged setting, RL with self-play seems to be the answer to planning for driving. impressive to see complex behavior learning without supervision and zero-shot generalization across benchmarks by GigaFlow. the next question is what about with perception, is it just a matter of computation?

“..unsere heimat wieder gross” I wish my German was not enough to understand this, what the hell is happening to Germany 🥹🥹

since Görkay came up with Track-On, we've been waiting for this day: Track with me, if it's just for the frame, maybe tomorrow, the occlusion will take you away! 🎵🎸😶‍🌫️ Happy to introduce our #ICLR2025 paper "Track-On: Transformer-based Online Point Tracking with Memory": kuis-ai.github.io/track_on/

open-source, open-weight, I'll take whichever because, unlike the ones who share nothing, a few at least make an effort and change the game for everyone. I expect the effect of open stuff like Cosmos to be huge on physical AI. we are already trying it. also downloaded the DeepSeek, we shall see 🤞

I recently started taking walks, you know, for health reasons. but somehow all these walks are directed towards a cafe where I can have a cake as a reward for my walk 😅

I want to work closely with the Turkish industry, I think we can help each other. they just seem to miss the point like they see us as charity*. funny I feel the same about them after our international collaborations. *or they see me as "just a girl standing in front of them, asking for money" 😂

I haven't made the full switch from X to bsky yet because of the Turkish community that is still active there but so many random accounts are rt'ing/following in the last few hours that it might be the time. I think Elon Musk is trying to create artificial traffic with bots. it's pathetic really.

AZ praised our work that I presented at VGG. I don’t know why I’m still such a fan girl, I’m certainly not proud, but it made me very happy 🥹 glad he is not on social media, pls don’t tell him 🤫😅

#ICLR decisions are out! we have a paper 🥳 I’m not even gonna try to play it cool 😅 our first ICLR paper, super happy for my student Görkay, and we are extremely lucky to work with Weidi on this project. stay tuned for a very simple online point tracking method with amazing results 🤩

I love that something as valuable as NeuroNCAP is open-source, so I'll advertise it: github.com/atonderski/n... a NeRF-based simulator trained on NuScenes to render new, unseen scenarios to stress-test e2e planners (UniAD and VAD) closed-loop in safety-critical scenarios.

Excited to share that today our paper recommender platform www.scholar-inbox.com has reached 20k users! We hope to reach 100k by the end of the year.. Lots of new features are being worked on currently and rolled out soon.

this is a very positive, nice way of thinking about progress but today I read a survey about 2024 papers I’ve been meaning to read 🥹 things are crazy fast right now, I have a hard time catching up with the existing work, let alone contribute my unique perspective 🥲

for a while, I was very scared that I’d become irrelevant because I chose to live in Turkey. with all the recent progress on the topics that I’d like to work on, I realize that I’d be irrelevant anywhere. well, that’s a relief :/

"Academia is the fifth circle of hell. It goes babies, musicals, magic, kale, academia." —Paul (Harrison Ford) on Shrinking. funnily enough, I like all these things on the list including kale (you just need to cook it right).

reading a book claiming that social media is no good unless it’s for common good. so sharing this, is it for common good? 🤔 also definitely no no for teenagers until high school at least, apparently it fries their brain like what Mars does to human body (example from the prologue of the same book).

in our group mtg today, someone said that they might be having a burnout. I asked other group members how they handle burnout, one of them said: “I just burn 🔥” 🥹😂

Efe, one of our interns last summer, implemented the ROACH expert on NAVSIM and apparently shared the code: github.com/efemantarci/... there is no documentation but it should be fairly easy to use in case you need an RL expert reaching the performance of TransFuser on NAVSIM.

will give a talk at VGG next week, very excited!! no matter how many talks you give, nothing like one’s homebase (well, one of homebases 😊)

another day, another code release! Nazir and Youssef did great work with likelihood ratio formulation for OoD segmentation. if you want to utilize DiNOv2 for OoD to achieve SoTA results with a pixel-level method, check it out! paper: arxiv.org/abs/2409.06424 code: github.com/NazirNayal8/...

code release for our ECCV paper CarFormer: github.com/Shamdan17/Ca... Shadi is an amazing coder, so this is a great starting point for privileged agents on CARLA. three versions to play with: 1. BEV quantized with a VQ-VAE 2. Attributes (PlanT) 3. BEV => slots (CarFormer)

two great workshops are coming your way this #CVPR: - Uncertainty Quantification for Computer Vision - Embodied Intelligence for Autonomous Systems on the Horizon kudos to collaborators and organizers 🎉

I feel seen 👀

do you also think that it’d be incredibly freeing to not care about awards and titles and just focus on good work? I sometimes dream of being much older, free of such worries and just do what I like, but in reality, the pressure to chase them -internal or external- increases with age so far.