Staff Research Scientist at @GoogleAI, previous: @MetaAI, @DartmouthCS, @NTUsg, @IllinoisCS. Opinions are mine.
Joined August 2008
- Tweets156
- Following244
- Followers1.7K
- Likes170
BudgetPix, our recent work, allows you to adapt any pixel-space diffusion into budget-flexible inference. It saves you more than half of the computation cost without a quality drop (noticable by human eyes). Great work by @ozgurkara99 and the team.
1/8 BudgetPix: one model, any compute budget.
Our pixel-space image generator adapts computation to image content at inference time. Same prompt, same noise, from 100% down to just 10% of the tokens.
→ Project Webpage: karaozgur.com/BudgetPix
Du Tran retweeted
📢Applications are now open for PhD visiting students @bristolcs @BristolUni in 2026 - DL 29 Jan.
Would you like to work with any of the Faculty working in Machine Learning and Computer Vision #mavi as part of our summer of research @ Bristol program?
uob-mavi.github.io/Summer@Ma…
Interested in the kind of projects you'll be working on? Check out these examples from previous interns:
1. SEAL: seal-lvu.github.io
2. StM: split-then-merge.github.io
We revisit the problem of video composition through the lens of GenAI.
🎥 Introducing Split-then-Merge: A new video composition framework!
This approach enables the composition of any foreground video with any background video.
Unlike conventional methods that rely on annotated datasets or handcrafted rules, Split-then-Merge (StM) splits a large unlabeled corpus of videos into dynamic foreground and background layers, then merges them to learn how dynamic subjects interact with diverse scenes.
Work done in collaboration with team members at @Google: Du Tran (@dutran) , Yujia Chen (@IssacCyj) , Prof. Ming-Hsuan Yang (@MingHsuanYang), Vincent Chu: and my advisor at UIUC (@siebelschool): Prof. James M. Rehg (@RehgJim).
I will be attending NeurIPS, San Diego and would be happy to chat more!
🔗Project Webpage: split-then-merge.github.io/
📄Paper: arxiv.org/abs/2511.20809
This new problem setup is both interesting and challenging. It demands a crucial balance: preserving the object's identity while simultaneously achieving a seamless and harmonious overall composition.
We are hosting the second edition of CV4DC at @ICCVConference. We look forward to great keynotes & talks and meeting you there in Hawaii on Oct 19.
We're excited to welcome you to CV4DC workshop at @ICCVConference this Oct 19, 2025 in Honolulu, Hawaii. We'll have a full-day workshop filled with 9 paper presentations and invited talks from leading researchers in CV.
Check out the program here: cv4dc.github.io/2025/schedul…
Our proceeding track is closed, but the non-proceeding track is still open for submission in case you want to submit. We had great memories last year in Hanoi (m.facebook.com/story.php?sto…). Hope to see you in Hawaii this December!
We’re thrilled to have received so many submissions to @cv4dcworkshop at #ICCV2025! 🎉
Huge thanks to everyone for your interest and contributions.
Our reviewers are hard at work—notifications coming soon. Stay tuned! 👀
CV4DC deadline is exteneded to July 2. Hope to see you this year at ICCV.
We’ve received great interest in @cv4dcworkshop
at @ICCVConference, and we’re giving you a bit more time to submit your work!
🗓️ New submission deadline: July 2, 2024 (AoE)
cv4dc.github.io/2025/call-fo…
Please help us spread the word!
#CV4DC #CV #ML #AIforGood #ICCV2025
Du Tran retweeted
We’ve received great interest in @cv4dcworkshop
at @ICCVConference, and we’re giving you a bit more time to submit your work!
🗓️ New submission deadline: July 2, 2024 (AoE)
cv4dc.github.io/2025/call-fo…
Please help us spread the word!
#CV4DC #CV #ML #AIforGood #ICCV2025
#ICCV2025 CV4DC workshop this year has 2 tracks:
1. Non-proceeding track, where you can submited accepted ICCV papers or previously published work, so your work can gain more visibility.
2. Proceeding track, is designed for unpublished / novel work.
Hope to see you in Hawaii!
We continue the second edition of CV4DC this year at #ICCV2025. CV4DC aims at fostering the community of CV/ML researchers from developing countries or reseachers interested in solving AI problems for developing countries. Please join us at ICCV25 in Hawaii this coming October!
Du Tran retweeted
🚀 Call for Papers! 🚀
We're excited to announce 2nd edition of CV4DC: Computer Vision for Developing Countries Workshop at #ICCV2025
Share your innovative research on computer vision addressing challenges in developing region
📅 Deadline: June 15, 2025
🔗 cv4dc.github.io/2025/
