Tengda Han

@tengda.bsky.social

Researcher at Google DeepMind. Computer vision and machine learning.

Thank @dimadamen.bsky.social for presenting our Orthogonal Optimizer! It’s a simple modification on standard optimizers for streaming video learning. We have code available at sites.google.com/view/orthogo...

Dima Damen@dimadamen.bsky.social · last yr.

Now @cvprconference.bsky.social poster session 3 #286 Our GoogleDeepMind paper: Learning from Streaming Video with Orthogonal Gradients As @tengda.bsky.social couldn’t make it for visa reasons, you’ll have the second best option of me presenting our work 😅 See you there #CVPR2025

Humans learn from one continuous visual stream, but large video models have to be trained on billions of web videos. We found that learning from such sequential streams is challenging for video models—and we introduce a family of "orthogonal optimizers" to bridge the gap!

We are looking for a student researcher to work on video understanding plus 3D, in Google DeepMind London. DM/Email me or pass it to someone if you feel it may be a good fit!

From an award candidate... to best paper #ACCV2024 Glad to share that "It's Just Another Day" received the top award at the conference. @bristoluni.bsky.social @ox.ac.uk This paper is worth reading :-) based on the reviewers, AC and awards committee. We thank them for their time and effort.

Bild
Dima Damen@dimadamen.bsky.social · 2y ago

Our #ACCV2024 Oral and Award Candidate It’s Just Another Day: Unique Video Captioning by Discriminitive Prompting now on ArXiv tobyperrett.github.io/its-just-ano... For challenging ego&timeloop movies, uniquely caption ev clip, including those near identical ones, w/out re-training captioning model