Pingchuan Ma

Video diffusion models learn motion indirectly through pixels. But motion itself is much lower-dimensional. We introduce 64× temporally compressed motion embeddings that directly capture scene dynamics. This enables efficient planning -> 10,000× faster than video models. 🧵👇

@CSProfKGD the size of the mac looks like a max pro ultra plus

Spending the day in Erlangen 🤗

Evening team building 🍻

now I know why they have rainbow on the car plate in Hawaii. #ICCV2025 #iccv25

A huge thank you to all my collaborators, @MingGui725184, @jo_schb, Xiaopei Yang, Yusong Li, @felix_m_krause, @GrebenkovaO, @vtaohu, and Björn Ommer @ CompVis/Ommer Lab ( @LMU_Muenchen)

Drop by to say hi or get a coffee together at #ICCV2025. I am also seeking full-time or internship opportunities 🧑🏻‍💻 starting June 2026 (apparently, this is quite important😏).

I'm happy to share that I’ll be presenting two first-authored papers at #ICCV2025 🌺 in Honolulu, together with @MingGui725184! 🏝️ (Thread 🧵👇)

🤔 What if you could generate an entire image using just one continuous token? 💡 It works if we leverage a self-supervised representation! Meet RepTok🦎: A generative model that encodes an image into a single continuous latent while keeping realism and semantics. 🧵👇

We cut the cost of training a diffusion model from months of rent to one night out. TREAD matches ImageNet performance of a DiT with 97% fewer A100 hours! No extra components. No extra losses. Training‑time only. Inference remains unchanged. Accepted at ICCV2025🌺

🤔 What happens when you poke a scene — and your model has to predict how the world moves in response? We built the Flow Poke Transformer (FPT) to model multi-modal scene dynamics from sparse interactions. It learns to predict the 𝘥𝘪𝘴𝘵𝘳𝘪𝘣𝘶𝘵𝘪𝘰𝘯 of motion itself 🧵👇

@gabriberton @CVPR @ICCVConference Hi Gabriele, do you know if they have already open the application site for the ICCV doctoral consortium this year? I could not find anything on the ICCV site.

I was at the doctoral consortium at @CVPR, what a great experience! By far the highest concentration of talent I've ever seen in a room. Plus the food was great :) I highly advise any PhD close to finishing (or who recently finished) to apply for it (coming at @ICCVConference)

Check out our new #CVPR20 paper on Transferring DensePose to Animals In collaboration with FAIR (V. khalidov, A. Vedaldi and @NataliaNeverova) 🌐 asanakoy.github.io/densepose-evol… ▶️ youtu.be/OU3Ayg_l4QM 📝asanakoy.github.io/densepose-evol…

Struggling with Metric Learning? Check out our new #iccv2019 @ICCV19 paper "MIC: Mining Interclass Characteristics for Improved Metric Learning" arxiv.org/abs/1909.11574 confusezius.github.io/metric-learnin…

Our style transfer on steroids at ICCV19! "Content and Style Disentanglement for Artistic Style Transfer" We learn subtle variations of styles and disentangle style from content Project page: bitly.com/disent-style Video: youtu.be/KN_WTcQBUsU #ICCV19

Last week I gave a talk at Heidelberg SIAM chapter "Identification of Humpback Whales using Deep Metric Learning". I talked about our recent CVPR'19 paper arxiv.org/abs/1906.05990 and about Humpback Whale Identification challenge at @kaggle. Slides: slides.com/asanakoy/metri…