mauricio delbracio
@2ptmvd
Research Scientist at Google. All opinions are my own. 🇺🇾
🤯🤯
5. Olga Souza singing "Rhythm of the Night" acapella, 1994
For better consistency and fewer artifacts in im2im diffusion models, consider Kernel Density Steering (KDS): It is inference time and training-free; uses an N-particle ensemble to derive an empirical density, & explicit mode-seeking to push samples to high-density regions 1/2
Excited to announce my work during @GoogleAI internship is now on arXiv! Grateful for my incredible hosts and collaborators: @2ptmvd, @docmilanfar, @KangfuM, Mojtaba Sahraee-Ardakan and @ukmlv. Please check out the paper here: [arxiv.org/abs/2507.05604].
For better consistency and fewer artifacts in im2im diffusion models, consider Kernel Density Steering (KDS): It is inference time and training-free; uses an N-particle ensemble to derive an empirical density, & explicit mode-seeking to push samples to high-density regions 1/2
+1000 this 👇🏼👇🏼
- Requiring each paper to contribute 3 reviews is fine. - Requiring each author to contribute 3 reviews is not.
⚠️ Deadline Extension for KHIPUx! 🗓️New deadline: July 14 📎 +info at: khipu.ai/khipux
KHIPU is more than a conference; it’s a vibrant community! With the spirit of collaboration at its core, we are proud to announce the launch of the second edition of the KHIPUx program, designed to expand this spirit across the heart of Latin America.
📢 We're hiring Research Scientists at @Google ! Join our team to develop cutting-edge generative models for imaging & computational photography, impacting millions of users. Apply now! +info: google.com/about/careers/…
I really enjoyed collaborating with @docmilanfar on this paper about denoisers! 🚀 Writing it helped a lot to deepen my understanding and distill my knowledge/intuition. For anyone new to the power of image denoisers, I think this is a great intro (though I might be biased!) 👇
Our paper is now formally published in: Phil. Trans. R. Soc. A.383:20240326 1/2
Announcing Ambient Diffusion Omni — a framework that uses synthetic, low-quality, and out-of-distribution data to improve diffusion models. State-of-the-art ImageNet performance. A strong text-to-image results in just 2 days on 8 GPUs. Filtering ❌ Clever data use ✅
Maybe the limit can be pushed a little bit down... like 25 --> 10?
#CVPR2025 paper stats
KHIPU is more than a conference; it’s a vibrant community! With the spirit of collaboration at its core, we are proud to announce the launch of the second edition of the KHIPUx program, designed to expand this spirit across the heart of Latin America.
KHIPU is more than a conference; it’s a vibrant community! With the spirit of collaboration at its core, we are proud to announce the launch of the second edition of the KHIPUx program, designed to expand this spirit across the heart of Latin America.
South America > Europe.
Muy lindo el show de Linkin Park, pero en mis libros, las cabras de shows fueron ellos.
"Soy inmigrante. Mis padres son refugiados de Chile. Huimos de una dictadura y tuve el privilegio de crecer en Estados Unidos tras recibir asilo en Dinamarca. Defiendo esas protecciones, siempre". - Pedro Pascal en Cannes. #Cannes2025
👀 notes by @gabrielpeyre
Optimal Transport for Machine Learners. arxiv.org/abs/2505.06589
If you submitted to ICML and TMLR as an author, where did you get higher review quality (grounded, fair, and helped improve your paper)? 𝐏𝐥𝐞𝐚𝐬𝐞 𝐝𝐞𝐜𝐨𝐮𝐩𝐥𝐞 𝐭𝐡𝐞 𝐮𝐥𝐭𝐢𝐦𝐚𝐭𝐞 𝐟𝐚𝐢𝐭𝐡 𝐨𝐟 𝐩𝐚𝐩𝐞𝐫 𝐟𝐫𝐨𝐦 𝐲𝐨𝐮𝐫 𝐣𝐮𝐝𝐠𝐞𝐦𝐞𝐧𝐭 𝐨𝐟 𝐪𝐮𝐚𝐥𝐢𝐭𝐲.
TMLR's bar is not lower. The car is drawn differently to remove subjectivity from the acceptance criteria. I have read many good TMLR papers and many disastrous ICML papers. In fact I think those disastrous ICML papers wouldn't stand a chance at TMLR.
TMLR defines it's bar for acceptance in a way I consider pretty clearly lower than conferences Completely agreed that peer review is dumb and noisy in many ways. But, like, I think that prestige is a real and meaningful phenomena that really affects careers
Happy to announce that our work, ShaRP, has been accepted to #ICML2025! 🎉 Huge thanks to my collaborators @ukmlv @2ptmvd @docmilanfar @WeijieGan1 @albert_peng_. Can't wait to see everyone in Vancouver!
🎉Thrilled to share our latest work “Stochastic Deep Restoration Priors for Imaging Inverse Problems”! Big thanks to my amazing collaborators @ukmlv @2ptmvd @docmilanfar @WeijieGan1 @albert_peng_. ⭑ Preprint: arxiv.org/abs/2410.02057. ⭑ Website: wustl-cig.github.io/sharpwww/.
📄 “Stochastic Deep Restoration Priors for Imaging Inverse Problems” was accepted to ICML 2025 (@icmlconf)! Joint work with @YuyangHu_666, @albert_peng_, @WeijieGan1, and colleagues from the Google’s Computational Imaging team, @2ptmvd and @docmilanfar. wustl-cig.github.io/sharpwww/
Nice work presenting @KangfuM !
At #ICLR2025? Experience the next step in computational photography today @ 3PM SGT at the Google booth where Kangfu Mei will demo a model that intelligently upscales images using estimated knowledge from diverse modalities like depth, segmentation, and text.…