Dmytro Mishkin 🇺🇦 (@ducha

Dmytro Mishkin 🇺🇦

21.8K posts

Dmytro Mishkin 🇺🇦

@ducha_aiki

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

Czechia

Joined July 2017

Pinned
Dmytro Mishkin 🇺🇦
@ducha_aiki
Jan 12
Update in the pinned post: The best way to support Ukraine is via donating to the trusted funds. A few of them @BackAndAlive - oldest and most trusted savelife.in.ua/en/donate-en/#… @sternenko - father of drones sternenkofund.org/en/donate Czech one: @DarPutinovi zbraneproukrajinu.cz
8K
Dmytro Mishkin 🇺🇦
@ducha_aiki
Feb 23, 2022
Dear friends. I realize that you followed me for the news in computer vision, image matching, machine learning. Not for political bullshit. But my country is in grave danger. It is not politics, it is life&death. I will continue to write about ML&CV. And also about 🇺🇦.
Dmytro Mishkin 🇺🇦
@ducha_aiki
Mar 4, 2021
I have submitted my PhD thesis, yay!
Dmytro Mishkin 🇺🇦
@ducha_aiki
Oct 5, 2021
Yesterday Facebook blackout shows why I am afraid of self-driving cars. Not because of deep learning: I believe we will solve robust perception at some point. But mega-centralization...one wrong update and millions of self-driving cars become a weapons of mass self-destruction.
Dmytro Mishkin 🇺🇦
@ducha_aiki
Feb 10, 2022
If you are interested in doing PhD in computer vision and/or robotics, consider CTU in Prague. We are bad in self-PR, but good in science. Also, Prague is a great and affordable city to live in.
Dmytro Mishkin 🇺🇦
@ducha_aiki
Oct 1, 2025
How Diffusion Models Memorize Juyeop Kim, Songkuk Kim, Jong-Seok Lee tl;dr: classifier-free-guidance is to blame arxiv.org/abs/2509.25705
19K
Dmytro Mishkin 🇺🇦
@ducha_aiki
Jun 9, 2020
That is 🔥 arxiv.org/abs/2006.04768… Linformer: Self-Attention with Linear Complexity Sinong Wang, Belinda Li, Madian Khabsa, Han Fang, Hao Ma
Dmytro Mishkin 🇺🇦
@ducha_aiki
Sep 20, 2021
Torch.manual_seed(3407) is all you need: On the influence of random seeds in deep learning architectures for computer vision @david_picard tl;dr: one can easily get +0.2 pp accuracy on ImageNet just by changing random seed. And much more on CIFAR10 arxiv.org/abs/2109.08203
Dmytro Mishkin 🇺🇦
@ducha_aiki
May 24, 2022
ReLU Fields: The Little Non-linearity That Could @AnimeshKarnewar, Tobias Ritschel, Oliver Wang, Niloy J. Mitra tl;dr: grid-based 3D representation + ReLU are almost NERFs, but much faster. arxiv.org/abs/2205.10824…
Dmytro Mishkin 🇺🇦
@ducha_aiki
Feb 26, 2022
❗️🇺🇦🇺🇦🇺🇦❗️ My friends have created Notion wiki with many ways how you can help #Ukraine - donation, petitions, housing, DDOS and many others #HelpUkraine Please, share. gskyiv.notion.site/gskyiv/Real-wa…
Dmytro Mishkin 🇺🇦
@ducha_aiki
Dec 27, 2022
Fully Differentiable RANSAC Tong Wei, @yash_patel2307 , Jiri Matas, @majti89 tl;dr: another (now working better?) differentiable RANSAC with clever pretraining (KL-div for inlier probabilities, then regular losses). Eval is good :) 1/2 arxiv.org/abs/2212.13185
61K
Dmytro Mishkin 🇺🇦
@ducha_aiki
Oct 2, 2024
DINO-CLIPosaur Is super crowded! #ECCV2024 @abursuc
72K
Dmytro Mishkin 🇺🇦
@ducha_aiki
Jul 31, 2021
Good practice, which I forget to do every other time. #OpenSource
Dmytro Mishkin 🇺🇦
@ducha_aiki
Oct 27, 2023
It seems that RoMA by @Parskatt almost solved WxBS. I have added it to eval colab here colab.research.google.com/drive/1yrCFyEo…
40K