Kosta Derpanis

38.2K posts

Kosta Derpanis banner
Kosta Derpanis

Kosta Derpanis

@CSProfKGD

#CS Assoc Prof @YorkUniversity, #ComputerVision Scientist Samsung, @VectorInst Affiliate, TPAMI AE, @ELLISforEurope Member #CVPR2026 #ECCV2026 Publicity Chair

Toronto, Ontario Katılım Ağustos 2015
199 Takip Edilen78.9K Takipçiler
Sabitlenmiş Tweet
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
Lecture slides for my "Introduction to #ComputerVision" and "#DeepLearning in Computer Vision" courses. 🆕 Gaussian Splatting 🆕 Flow Matching 🆕 Sparse Autoencoders The included videos do not contain voiceovers yet, planned for a future revision.
Kosta Derpanis tweet mediaKosta Derpanis tweet media
English
8
88
571
40.8K
Kosta Derpanis retweetledi
Mariya I. Vasileva
Mariya I. Vasileva@mariyaivasileva·
Research shouldn’t turn into a deadline-driven content creation pipeline. Good research in my experience looks nothing like this. You think deeply about a problem, approach it from multiple angles, revise the direction repeatedly, solicit peer feedback, and refine. By iteration N, the paper usually looks quite different from the original idea, and what you thought was “the paper” might end up as just one section of the final result. Writing — specifically paper positioning, findings framing against prior work, visualization of main results, has always taken me even longer to solidify. Plus, peer review is already strained beyond capacity, this practice just makes it low-signal on both sides.
Amit LeVi@AmitLeViAI

Such a great evening to start a brand new research for NeurIPS in 3.5 days.🧘‍♂️ Day 1: planning. Night 1: running experiments and sending the abstract. Day 2: reading results fighting with Claude, and sending again. Night 2: sleep (optional). Day 3: opening Codex, and finally, write the pape in parallel. Night 3: resolving the “beef” with Claude (temporary peace) and going to sleep. Day 4: final reading, last-minute fixes, submission then some relaxation, maybe a beach walk. I’ll keep you posted on the results. This will be my only single-author paper, so I can’t hide behind other submissions if it gets rejected 😅

English
3
12
115
8.3K
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
Updating your method with a reviewer’s suggestions 🙄
English
1
0
3
449
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
@AmitLeViAI @mariyaivasileva You seem to have a lot of time to both do research in 3 days and keep up with the discussion you’ve stirred up at the time 🤓😅
English
1
0
3
68
Amit LeVi
Amit LeVi@AmitLeViAI·
@mariyaivasileva When I publish this paper, people will ask, ‘How did he do this in three days?
English
1
0
1
406
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
The state of many ML/CV rebuttals 😅
Kosta Derpanis tweet media
English
2
0
10
1.2K
Kosta Derpanis retweetledi
Kostas Daniilidis
Kostas Daniilidis@KostasPenn·
With deep sadness I share the passing of Theo Pavlidis (1934-2026) pioneer of pattern recognition and image analysis, K.S. Fu Prize laureate, IEEE Life Fellow, author of foundational texts in pattern recognition that shaped a generation. Many in the computer vision community are academic descendants including @CSProfKGD.
Kostas Daniilidis tweet media
English
1
8
29
1.7K
Gabriele Berton
Gabriele Berton@gabriberton·
While people argue about PyTorch vs JAX, I remember the first deep learning codebase I had to deal with It was in MATLAB Yes, MATLAB was also used for deep learning before the big frameworks took over It was for NetVLAD (2015)
Gabriele Berton tweet media
English
11
2
48
4.1K
Kosta Derpanis retweetledi
Roei Herzig
Roei Herzig@roeiherzig·
If you can complete a research paper in 3.5 days, it is probably not a very high-quality one. Good research takes time; there are no real shortcuts.
Amit LeVi@AmitLeViAI

Such a great evening to start a brand new research for NeurIPS in 3.5 days.🧘‍♂️ Day 1: planning. Night 1: running experiments and sending the abstract. Day 2: reading results fighting with Claude, and sending again. Night 2: sleep (optional). Day 3: opening Codex, and finally, write the pape in parallel. Night 3: resolving the “beef” with Claude (temporary peace) and going to sleep. Day 4: final reading, last-minute fixes, submission then some relaxation, maybe a beach walk. I’ll keep you posted on the results. This will be my only single-author paper, so I can’t hide behind other submissions if it gets rejected 😅

English
5
1
91
14.5K
Kosta Derpanis retweetledi
Phillip Isola
Phillip Isola@phillip_isola·
@HowToAI_ Thanks for sharing! I wouldn't say we "proved" this :). There is lots of evidence for convergence and lots against it. To me the interesting next question is *which structures* converge and which do not? I think more converges than we used to think, but not everything.
English
4
3
69
3.6K
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
Next stop 🤗 Last leg of my sabbatical.
Kosta Derpanis tweet media
English
0
0
12
943
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
#ECCV2026 reviews didn’t go your way? Accept is still possible. Focus on the substance of the reviews rather than fixate on the ratings. Thoughtful, targeted rebuttals, combined with some luck having engaged reviewers and ACs, can make the difference.
English
2
2
33
3K
Kosta Derpanis
Kosta Derpanis@CSProfKGD·
Saturday date night. Critics hate it but audiences ❤️ it 🤔
Kosta Derpanis tweet mediaKosta Derpanis tweet media
English
1
0
6
1.8K

Keşfet

@AmitLeViAI @mariyaivasileva @jponttuset @UHN @gabriberton @HowToAI_ @elonmusk @BarackObama