jo.schb (@jo_schb) header
jo.schb (@jo_schb) profile picture

jo.schb

@jo_schb

PhD student @ CompVis Group, LMU Munich

Joined January 2021
  • Tweets68
  • Following610
  • Followers319
Diffusion models treat every part of an image equally. → Same number of steps. Same compute. But images aren’t uniform. 🤔 Some regions are easy, others are hard. So why force the model to treat them the same? 🧵
11
93
9
657
83,145
⚡️Few-step diffusion without a teacher, self-distillation or JVPs. iDDM: a single model trained from scratch in one stage. ImageNet-256: 4.48 FID at 4 steps and 2.38 at 50, from the same checkpoint 🧵 (1/6)
Image from Tommaso Martorella (@_tommymarto_)
5
36
1
194
12,275
jo.schb retweeted
Jannik (@Ragor_) profile picture
Jannik@Ragor_Jul 29
People often think of prediction as choosing a likely outcome. But that’s not how the world works! Given partial information, the future is a distribution of possibilities. And more knowledge collapses the distribution. We built a representation that models exactly that 🧵
1
8
1
15
1,477
❤️
Kosta Derpanis (@CSProfKGD) profile pictureKosta Derpanis
@CSProfKGDJul 27
Afternoon reading
Image from Kosta Derpanis (@CSProfKGD)
1
7
3,123
Most representation learning methods produce generic embeddings that try to preserve everything in an image. We introduce VICIS: a way to use example sets to define a tailored embedding space for what matters. This is useful whenever the desired visual signal is easier to show than to describe. Code & weights available now 🧵👇
Image from Kolja Bauer (@KoljaBauer)
1
15
19
1,347
I am at @CVPR in Denver this week. If you’re around and want to chat, feel free to send me a DM or stop by one of the two posters I’ll be presenting👇
Image from jo.schb (@jo_schb)
2
6
533
Denoising, Fast and Slow: Difficulty-Aware Adaptive Sampling for Image Generation 🕐Sun 7th, 3:30 PM - 5:30 PM 📍ExHall A 658 teapawt.net/jo_schb/status/2048765…
jo.schb (@jo_schb) profile picturejo.schb@jo_schbApr 27
Diffusion models treat every part of an image equally. → Same number of steps. Same compute. But images aren’t uniform. 🤔 Some regions are easy, others are hard. So why force the model to treat them the same? 🧵
1
2
280
Probabilistic Precipitation Nowcasting with Rectified Flow Transformers 🕐Sat 6th, 4:45 - 6:45 PM 📍ExHall A & F 401 teapawt.net/jo_schb/status/2062333…
jo.schb (@jo_schb) profile picturejo.schb@jo_schbJun 4
⚠️ Standard first stages are not sufficient for safety-critical applications! The most extreme weather events are often the hardest to decode. One latent → many plausible reconstructions Deterministic decoders hide that uncertainty. Meet FREUD 🧵👇
1
2
237
⚠️ Standard first stages are not sufficient for safety-critical applications! The most extreme weather events are often the hardest to decode. One latent → many plausible reconstructions Deterministic decoders hide that uncertainty. Meet FREUD 🧵👇
2
7
2
10
1,450
more replies
For more information: Project page: compvis.github.io/weather-rf Paper: arxiv.org/abs/2605.31204 Code: github.com/CompVis/weather-r…

Probabilistic Precipitation Nowcasting with Rectified Flow Transformers

FREUD enables uncertainty-aware and efficient precipitation nowcasting in latent space.

compvis.github.io
1
3
108
Joint co-first work with @Ragor_ , and collaborators @rmsnorm, Timy Phan. Supervised by Björn Ommer @ CompVis Group LMU Munich
3
141
The internet is full of video. So why can't novel view synthesis just scale on it? Real-world video is simultaneously unposed, messy, and dynamic, breaking self-supervised NVS. We fixed that. RayDer learns static-scene NVS from dynamic internet video, scaling like an LLM. A🧵
8
38
4
184
28,171
Takeaway 🚀 • Diffusion shouldn’t treat all regions equally • Patch-wise timesteps improve performance, if done right • Allocating compute where it matters gives further gains Project Page: compvis.github.io/patch-forc… arXiv: arxiv.org/abs/2604.19141 Code: github.com/CompVis/patch-for…
1
3
46
3,007
Joint work with @MingGui725184, @Yusong53080064 , @PingchuanMa4, @felix_m_krause, and Björn Ommer! 🫶
1
9
1,662
Load more