Danush

PhD student @TSO_lab

[Danush @kumar_danush]

📢 We introduce SurgicaL-CD, a diffusion model to generate realistic, labeled surgical images in one step

🎉 I'm excited to share that our work is accepted at the ECCV 2024 workshop on Synthetic Data for Computer Vision
Paper : https://t.co/vErSfHDxPi
1/5🧵

[Danush @kumar_danush]

@arnie_hacker Reading mp4 files directly during training worked better for me. You can try using decord or torchvideo for loading or decoding the video directly on the GPU. But I'm not sure for hour long videos. Create smaller snippets with ffmpeg.

[Danush @kumar_danush]

@yash2kant @igilitschenski @UofTCompSci @Snapchat Congrats @yash2kant 🎉

[Danush @kumar_danush]

@k44yej @MicrosoftAI Happy to chat !

[Danush @kumar_danush]

@HAOTIANCUI1 @Xaira_Thera @BoWang87 Hey !! Happy to chat about foundational models

[Danush @kumar_danush]

How to make sure this phenomenon is not happening? Maybe this does not happen at a larger scale ? Happy to hear thoughts from diffusion folks
#diffusionmodels #NeurIPS2025

[Danush @kumar_danush]

FID for generative models : If we monitor FID during DM training and expect it to go down over longer training steps, what are we measuring here 🤔

[Danush @kumar_danush]

What if the model merely starts to replicate the data from the training set ? This could also lead to lower FID 😉

[Danush @kumar_danush]

@YuyangW95 Would love to chat about diffusion models !! Your DMs are closed.

[Danush @kumar_danush]

@sumukhaithal6 @MistralAI Hi Sumukh !! Happy to chat at NeurIPS

[Danush @kumar_danush]

@borruell @iconicgamesio Hey !! I would be happy to chat about world models over coffee tomorrow:)

[Danush @kumar_danush]

@_vztu @NeurIPSConf Would be happy to talk about video generative models

[Danush @kumar_danush]

@BoWang87 @Xaira_Thera Looking forward to meeting at the party !!

[Danush @kumar_danush]

@BDuisterhof I've headed to NeurIPS as well. Happy to meet and chat !

[Danush @kumar_danush]

@BoWang87 @Xaira_Thera That's great news. I'm interested in meeting you at NeurIPS. I would be happy to chat about the roles.

[Danush @kumar_danush]

@gowthami_s I resonate this with people who can sell themselves well although they might not have what's needed.

[Shawn Shen @shawnshenjx]

I’m Shawn, founder of https://t.co/DLaNyXWuUn, former researcher at Meta and CS PhD at University of Cambridge.

Today we’re launching https://t.co/DLaNyXWuUn: we built the world’s first Large Visual Memory Model - to give AI human-like visual memories.

Why visual memory?

AI to date is chat-based, which has great applications, but humans are not only chat based. Humans have visual-memories, and are visually driven. https://t.co/DLaNyXWuUn enables AI agents, softwares, and even robots to see and remember the world the way humans do.

We just raised an $8m seed round led by @susaventures to build the visual-memory layer for AI.

Why is https://t.co/DLaNyXWuUn so groundbreaking? 👇

[Danush @kumar_danush]

@cloneofsimo Any quick interesting insights from this work ?

[Danush @kumar_danush]

@cloneofsimo Certain form of scaling was done in this exciting work https://t.co/V73JpWVxRv

[Danush @kumar_danush]

@CSProfKGD @KostasPenn Are there more photos of the slides ?

[Danush @kumar_danush]

@yuvalalaluf @runwayml Congrats 🎉