🚨 New paper out! "VisualOverload: Probing Visual Understanding of VLMs in Really Dense Scenes" 👉 arxiv.org/abs/2509.25339 We test 37 VLMs on 2,700+ VQA questions about dense scenes. Findings: even top models fumble badly—<20% on the hardest split and key failure modes in counting, OCR & consistency.

How does AI interact with culture? We’re thrilled to have Dr. Maria Antoniak (@mariaa.bsky.social) join us at FGVC! Her interdisciplinary work on AI in the humanities brings a new perspective to our workshop: the fine-grained categorization of the intangible parts of culture. See you at #CVPR2026!

🚨 Updated Deadline Alert! We have extended the deadline for the Proceedings from Feb 27th to Mar 3rd. #CVPR #CV #AI

🚨 Reminder: The Proceedings Track deadline is February 27. Don't miss out on this chance to share your research 😁 We're very excited to see what you've been working on! #CVPR2026 #AI #ML

We are excited to announce that we are co-hosting the AnimalCLEF25 challenge with LifeCLEF at #CVPR2026! 🐾 Individual animal re-identification is a major challenge in conservation. Help us build evidence-based conservation tools that work! Find out more: www.kaggle.com/competitions...

Man, the last year has really kicked my ass and this CVPR reject is such a painful one. I just needed one more paper accepted to complete my thesis so I can finally move on, and this year has just seen reject after reject after reject. I'm so exhausted. How does anyone keep working through this?

We are jointly hosting the FathomNet challenge at CVPR this year! If you're interested in solving problems at the frontier of Positive-Unlabeled learning in Marine environments, dive in! 🏊‍♀️ 🦑 32 marine classes (from squat lobsters to benthic worm) 📉 6.4k training images (incomplete labels) #CVPR

Waiting for CVPR results after the reviewers cooked me.

SAM 3D sometimes has cute/funny failure modes under occlusion for complex object shapes. Who is this little gremlin? Why did the model look at this box of kittens and think "that there in the back is a gremlin for sure"?

FGVC13 submissions are OPEN NOW! Whether you have new or existing work, we have a track for you: 1. Proceedings: Due Feb 27 2. Non-archival: Due April 3 3. Nectar (Already published): Due April 30 Full details: sites.google.com/view/fgvc13/... #AI #CVPR2026 #MachineLearning #DeepLearning

I'm really excited to join the organizing team for the 13th FGVC workshop! If you're into the nitty-gritty of distinguishing between similar bird species, spotting rare diseases, or even generating very specific characters, you should definitely check this out. #CVPR #CVPR2026 #AI #ML

AGI will be achieved when an AI can successfully take a duvet cover out of the laundry and put the cover on a blanket. That's it.

One idea for improving reviews at conferences is to assign one reviewer as a champion by default. Their role is to find every possible reason to accept the paper. I find reviewers often spend most of their time focusing in on what is bad in a paper. This is useful! But more often than not

Now that the #AAAI2026 discussion is mostly complete, I want to give my final assessment of the AI summary from the perspective of a reviewer: 1. The positives: It functioned like a literature review + found minor errors and inconsistencies in the paper. Really useful as a high-level sanity check.

This is an VERY important point, and I wonder how much knowledge we have lost because an interesting research path is cutoff due to early results not outperforming big models on some benchmarks. We are "expecting skyscrapers to learn to fly if we build them tall enough", while punishing other ideas.

Want to evaluate open-world object counting in a fine-grained setting? We’re excited to release the LookAlikes dataset, which is a test-set-only benchmark where images contain objects from multiple visually or semantically similar categories! #ComputerVision #DeepLearning #AI #Counting #OpenDataset

Why are there only 2 major CV conferences a year? If I want to submit a deep learning paper, there is NeurIPS/ICLR/ICML/AAAI. For a CV paper, I only have CVPR and ECCV|ICCV (Nov/Mar deadlines). This year there were 10K CV papers submitted to AAAI and 11K to ICCV. There should be a July CV deadline

GPT5 still can't solve this visual puzzle, which is interesting!

I guess we know what's left for GPT6

I've been reading a bit of Byung-Chul Han, recently, and I'm starting to think the male loneliness epidemic is just a byproduct of digital ecosystems destroying our ability to form coherent narratives about our lives. Dating is reduced to listing of algorithm friendly facts. We convert ourselves to

Paper idea: Adversarial images but for people. When shown this image of a real sand sculpture, organic intelligence tends to identify it as fake.

On deep learning and late capitalism: I was listening to a talk by Mark Fisher on the "Slow Cancellation of the Future" and he makes this point that the internet has made the past available in a decontextualized way where it becomes an aesthetic choice rather than anything coherent and organized.

Damn 😥

Elevated heart rate? Focus in shambles? Ah, yes, it's decision day for #ICCV2025 🤢🥶

The predictive power of the 20 year nostalgia cycle is unparalleled.

o4-mini-high doesn't know about "Loss"

This is an interesting visual puzzle, which o4-mini-high fails to solve. It never considers the orientation of the letters with respect to each other and thus that there must be one letter that exists twice on the cube.

Did gpt-4o get a bunk update recently? I gave it this riddle and it fell through the floor.

Now you're thinking strategically and hitting on one of the core contentions of the review process! Let's dig into it --

The USA is engaging in religiously-motivated organ harvesting targeting woman. Like, just out in the open.

When your 1-page rebuttal gradually starts feeling like the optimal packing of 17 squares 😵‍💫 #ICCV2025

Asking ChatGPT to convert the children's drawing (left) to an anime character (middle) vs. a human attempt (right).

ChatGPT has gotten substantially worse as a critic. It really wants to tell you how good and smart you are. Excuse me but work is NOT good and I'm not smart enough to notice why. Please do your JOB. Go to the ICLR openreview and find the lowest rated paper you can. Ask ChatGPT is it would accept the

There’s a pattern in deep learning I think of as “first to a dead-end.” Defining a novel problem often gets you a flashy paper, but follow-ups become hard to justify without being labeled incremental. So the problem is quietly abandoned. It’s a product of an oversaturated field, where it's safer to

I waited 2 hours to vote in the Vancouver byelection yesterday. I'm absolutely ecstatic to see a landslide for socialist and progressive candidates Sean Orr and Lucy Maloney.

Why are the ICCV submissions in my review pile so good 😭 you all got me sweating for my submission.

What episode of Severance is this?

I wish bluesky was a consumer co-op and membership gave you verification as well as the right to participate in the AGM and vote on things like the board. I would be dumping money into that.

Full lineup: Putin, Rodrigo Duterte, an LLM trained exclusively on Mein Kampf, the restless shade that lingers in the mirrors on Epstein island, and a picture of a slur.

I looked into the tree trunk detection dataset a few years ago when I was trying to find challenging datasets for object counting eval. But I kind of forgot about trees as a subject for AI until l saw a poster about making 3D trees with diffusion at ECCV. I'm interested to see where this paper goes!

I do hate when Trudeau sort of throws Mexico under the bus when appealing to Trump. Trudeau, who has abysmal support, could learn a lot from Claudia Sheinbaum, who is one of the most popular politicians.

P/E ratio is a 131, so lots of room for downside! 😁

Key takeaway from DOGE: There’s less government fraud/waste than expected.

I'm currently reading "An American Utopia: Dual Power and the Universal Army" by Jameson and I can't help but look at Chollet's comments through that lens -- here, we have an attempt to build up European military capacity through private enterprise and the neoliberal model and I frankly just can't

Yes it can. Maybe not forever, but the USA is a country where whole generations of Black people were born into and died under a brutal system of slavery. We must treat tyranny as the existential threat that it is.

Slava Ukraini 🇺🇦 and Free Palestine 🇵🇸

This morning I had a paper rejected from CVPR, which I really really liked. Boy oh boy it stings like nothing else! The reviews skewed positive as well, so I had high hopes ❤️‍🩹 But to everyone who got their papers through, congratulations 🎉

Personally, I welcome DOGE. Now I can promote my new extremely efficient cancer treatment -- first, you completely incinerate the person, technically destroying the cancer. Then you give Grok all of the patients social media posts and send their family an AI copy of their loved one.

When a billionaire tells you that the revolution involves AI memes, he wants you to forget that this is a class war.