Sign in

Dmytro Mishkin

@ducha-aiki.bsky.social
2.7K followers 164 following 1.6K posts

Marrying classical CV and Deep Learning. I do things, which work, rather than being novel, but not working. dmytro.ai

PostsRepliesMedia
Dmytro Mishkin @ducha-aiki.bsky.social · 16h
Accepted to #WACV2027
080
Dmytro Mishkin @ducha-aiki.bsky.social · 12/10/2026
Yes, my agents also brought it it me and it is beautiful
010
Dmytro Mishkin @ducha-aiki.bsky.social · 12/10/2026
To have accuracy as "circle size" now in my top of misleading plot examples. Not to mention, that star doesn't obey even this strange idea
070
Dmytro Mishkin @ducha-aiki.bsky.social · 10/10/2026
Well. Russia should be banned from everywhere.
120
Dmytro Mishkin @ducha-aiki.bsky.social · 07/10/2026
Nobody can steal your software if you giving it for everyone for free.
static.klipy.com
Man Tapping Head: Think About It Meme
ALT: Man Tapping Head: Think About It Meme
0120
Dmytro Mishkin @ducha-aiki.bsky.social · 07/10/2026
Agents finally brought the promise of all data labeling companies failed to deliver - you can label a little of data, new model/alg is trained on them and helps you more and more. The labeling is becoming adaptive as well. I love this
080
Reposted by Dmytro Mishkin
Christian Wolf @chriswolfvision.bsky.social · 06/10/2026
Having fun with Blender ... Distilling context transformers into recurrent transformers: arxiv.org/abs/2606.21562 @weinzaepfelp.bsky.social @mbsariyildiz.bsky.social
1225
Dmytro Mishkin @ducha-aiki.bsky.social · 06/10/2026
The best and the worst work for the agents is speed optimizations. Best - because it is well defined and easy to climb leaderboard. Worst - because you cannot use agent swarm, as the timings would be unreliable.
070
Dmytro Mishkin @ducha-aiki.bsky.social · 06/10/2026
Some work on similar, but for algorithmic case, when you kind of want to do it. arxiv.org/pdf/2503.13433 Earlier work from ours, arxiv.org/pdf/2003.01587 The issue is not tuning on test alone - it is tuning for YOUR method, crappy baseline of other - every line crossing is opportunity to cheat.
120
Reposted by Dmytro Mishkin
Jan van Gemert @jvgemert.bsky.social · 06/10/2026
In addition to the research findings, this is the type of research I really enjoyed doing 😊 "On Hyperparameter Tuning on the Test Set" arxiv.org/abs/2610.05902
arxiv.org
On Hyperparameter Tuning on the Test Set
"Don't tune hyperparameters on the test set" is often stated in machine learning textbooks. Violating it is considered a cardinal sin that produces misleadingly optimistic results, corrupts benchmark ...
2234
Reposted by Dmytro Mishkin
arXiv @arxiv.bsky.social · 01/10/2026
arXiv has updated our policy on rate limiting for all submitters. This update was made to fairly distribute moderator time & support the arXiv community of staff, volunteers, readers & authors. Please read our announcement to learn more: blog.arxiv.org/2026/10/01/updated-r…
216874
Reposted by Dmytro Mishkin
Maria Antoniak @mariaa.bsky.social · 02/10/2026
One of the aspects of industry chatbots that scares me the most. I have refused to work on persuasion enhancement since early in my PhD.
2445
Dmytro Mishkin @ducha-aiki.bsky.social · 29/09/2026
Oh yes, great work
000
Reposted by Dmytro Mishkin
International Conference on 3D Vision @3dvconf.bsky.social · 29/09/2026
🏆 The 3DV Citations Grand Prix! P2, with 3337 citations. "Matterport3D: Learning from RGB-D Data in Indoor Environments". arxiv.org/abs/1709.06158 Shout out to Angel Chang, Angela Dai, Thomas Funkhouser, Maciej Halber, Matthias Nießner, Manolis Savva, Shuran Song, Andy Zeng, and Yinda Zhang!
283
Dmytro Mishkin @ducha-aiki.bsky.social · 26/09/2026
Is it still true: is there trigonometry in US high schools
000
Dmytro Mishkin @ducha-aiki.bsky.social · 25/09/2026
Have you heard about CMT leaks?
120
Reposted by Dmytro Mishkin
Kosta Derpanis @csprofkgd.bsky.social · 23/09/2026
Visualization of the day: Slices through visual spacetime Interactive webpage: csprofkgd.github.io/spacetime-vo...
3215
Reposted by Dmytro Mishkin
erotemic.bsky.social @erotemic.bsky.social · 22/09/2026
TIL - unlike Python - bash reads scripts incrementally. ``` echo "Starting..." sleep 30 # Change this before the sleep finishes, and the new message will be printed. echo "Finished!" ```
172
Reposted by Dmytro Mishkin
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 21/09/2026
Example questions: (1) What papers look like after AI (2) What is would look like for a positive society-wide benefit from AI (3) What job training looks like if tasks become easier Trying to work through my own vision and it's nice to assemble others
4242
Reposted by Dmytro Mishkin
Thomas Dietterich @tdietterich.bsky.social · 20/09/2026
A lesson from decades of work on automation is that autonomous systems require continual monitoring and human oversight. System designers never(*) succeed in anticipating all of the ways that a system can fail.
13917
Dmytro Mishkin @ducha-aiki.bsky.social · 20/09/2026
Amazing little feature on bsky -- automatic indicator of number tweets in the thread.
1111
Dmytro Mishkin @ducha-aiki.bsky.social · 20/09/2026
Tons of them
000
Dmytro Mishkin @ducha-aiki.bsky.social · 20/09/2026
New, not so slop AI pop science video on YouTube. Kind of reasonable topic and not terrible delivery, didn’t even saw to “ai content” stamp. Next video topic helped me though: “Why we are eating tilapia and not carp?” WTF you mean not eating carp? Will stick to Veritasium
130
Dmytro Mishkin @ducha-aiki.bsky.social · 18/09/2026
Depends on other losses, often are helpful and easier to learn than 3D points themselves.
011
Dmytro Mishkin @ducha-aiki.bsky.social · 18/09/2026
@jacklangerman.bsky.social you may need to consider working at Naver…
110
Dmytro Mishkin @ducha-aiki.bsky.social · 18/09/2026
It is always tempting to add "consistency loss term" in 3D in one way or another. Do not do it. If it helps, then your data is too small, or you are predicting disconnected outputs. Just clf/regress vs ground truth, and do it long enough.
190
Dmytro Mishkin @ducha-aiki.bsky.social · 17/09/2026
I still skim through the ScholarInbox, and still read papers, but like once is a week and usually from people I know. Also I start to read old papers. VidMap is ETH paper, I have actually read it twice, didn't get correctly, talked to authors, now maybe I do understand. :)
131
Reposted by Dmytro Mishkin
Noa Garcia @noagarciad.bsky.social · 17/09/2026
our latest work on instance-level recognition by learning from generated images is out! with @yankunwu.bsky.social, Zakaria Laskar, @gkordo.bsky.social and @gtolias.bsky.social
0133
Dmytro Mishkin @ducha-aiki.bsky.social · 17/09/2026
I am waiting on New Zealand joining the list)
010
Reposted by Dmytro Mishkin
Bill Psomas @billpsomas.bsky.social · 16/09/2026
🔥 Composed image retrieval, as it should be. 📸 i-CIR (NeurIPS 2025): a photo says which particular object, a text says how it should look. 🔍︎ Explore every composed query in our new dataset browser: vrg.fel.cvut.cz/icir/ More to come... #ComputerVision #ImageRetrieval #AI #MultimodalSearch
vrg.fel.cvut.cz
i-CIR: Instance-Level Composed Image Retrieval (NeurIPS 2025)
A photo says which object, a text says how it should look. 202 instances, 1,883 composed queries, 752K curated hard negatives as hard as 40M random distractors, and an interactive dataset browser.
161
Dmytro Mishkin @ducha-aiki.bsky.social · 17/09/2026
Sadly, I don’t monitor arXiv anymore. I don’t even read Scholar Inbox suggested papers, although I am still going to read VidMap carefully. If you subscribed to me for the arXiv feed only, probably unsubscribe.
1100
Dmytro Mishkin @ducha-aiki.bsky.social · 16/09/2026
YES
020
Reposted by Dmytro Mishkin
Gautam Kamath @gautamkamath.com · 16/09/2026
In most practical cases this quickly happens automatically due to the quota system (all submissions count, including desk rejections) medium.com/@TmlrOrg/ann...
medium.com
Annual Author Submission Quotas for TMLR
TMLR submissions have grown substantially in recent times (Figure 1), tripling in the past one year. Among these, the number of…
0162
Reposted by Dmytro Mishkin
Gautam Kamath @gautamkamath.com · 16/09/2026
Nihar Shah did a heroic experiment for TMLR: he spent 20-25 hours over two weeks interviewing authors of seemingly low-quality submissions about their own papers. He confirmed what we all suspected: people submitting these papers have *no idea* what is going on in them.
5293117
Dmytro Mishkin @ducha-aiki.bsky.social · 16/09/2026
Amazing post
020
Reposted by Dmytro Mishkin
jonathan zong @jonathanzong.com · 16/09/2026
gotta keep tapping the sign peoples-things.ghost.io/why-do-peopl...
peoples-things.ghost.io
Against Narcissistic-Sociopathic Technology Studies, or Why Do People *USE* Technologies?
Why and how do people USE technologies? This question should be at the center of any thinking about human life with technology. The ways in which people use technologies, after all, determines the so...
2112
Reposted by Dmytro Mishkin
Tetiana Martyniuk🇺🇦@ ICIP’26🇫🇮 @t-martyniuk.bsky.social · 16/09/2026
(1/n) Can older lidar SSC models catch up with recent SOTA just by getting better inputs? At #ICIP2026 today and in our new journal extension, we add two cheap priors to existing networks, with no redesign. 📍 Poster #2752, Park Hall, 14:00–16:00 🇫🇮 🧵👇
182
Reposted by Dmytro Mishkin
David Picard @davidpicard.eurosky.social · 16/09/2026
Here is a thread of all the good reasons to use outlook! 🧵👇
3222
Reposted by Dmytro Mishkin
Philippe Weinzaepfel @weinzaepfelp.bsky.social · 13/09/2026
When were #ECCV2026 papers available on arXiv? 👇
2116
Reposted by Dmytro Mishkin
Tetiana Martyniuk🇺🇦@ ICIP’26🇫🇮 @t-martyniuk.bsky.social · 13/09/2026
#ICIP2026 First impressions: the venue is amazing! If you’re attending and wanna talk all things 3D scene understanding - hit me up on the conference app!
092
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
Thank you!
010
Reposted by Dmytro Mishkin
Nicolas Audebert @nshaud.bsky.social · 12/09/2026
Yeah the booklet would have been a lot better. Conference apps can be good, but Cvent is not. It's slow, the UI is terrible and it crashes on my phone when listing sessions. And the ECCV website is not better. 🤷🏻‍♂️
021
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
That’s a regret, ok. arxiv.org/html/2304.14...
arxiv.org
Regret Optimal Control for Uncertain Stochastic Systems
000
Reposted by Dmytro Mishkin
Christian Wolf @chriswolfvision.bsky.social · 12/09/2026
ECCV Poster #305 at 3pm (now): A scalar per patch from pre-trained ViTs enables fast moving navigation in the real world @steevenj7.bsky.social, L. Antsfeld, myself (I will present) - Visual enc distilled from heterog. teachers - Attention proj - Pretrain Policy w Lidar input => remove Lidar
1154
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
A robot moves hook from A to B and we measure battery drain (assuming perfect operation). Hook catches 50kg bag of rice. Stupid robot continues to pull hook and bag, while clever one untangles and proceeds w/o bag. Is there metric/formula for taking stupidity into account?
120
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
#ECCV2026 was an amazing event. Good people, good program, good food, good location. Thanks to all orgs, chairs and reviewers. @eccv.bsky.social P.S. There were small issues I complained about, but they are minor and don’t change my strong accept recommendation.
080
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
I actually miss the confidence booklet in paper form, where you can immediately see the paper names/schedule and make notes. Cvent app is way worse that it, almost useless for papers. Workshops were fine though @eccv.bsky.social
140
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
I am glad that many bad papers also have bad posters, it really helps filtering. And often that is not a lack of skill, but rather lack of care
040
Dmytro Mishkin @ducha-aiki.bsky.social · 12/09/2026
What were the main take aways? I missed the panel :(
100