/AI12h ago

CVPR Paper Pretrains 3D Lidar Models from Video Without Real Scans

048873.4K
Original post
Andrei Bursuc @CVPR@abursuc#1577inAI

Find out how we get strong Lidar encoders without Lidar data. We are in poster session 6 stand #270 #cvpr2026 @y_m_asano @FragileGoodwill @gatheluck

2:49 PM · Jun 7, 2026 · 1.8K Views
Sentiment
Sentiment building, check back later.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS1.5KBOOKMARKS3LIKES37RETWEETS5
Yuki@y_m_asano

If you only take a look at one poster in this last session come to 270!! 💪 We learn 3D representations from raw web videos via video-generated point clouds thanks to pi3! + we introduce a new better SSL loss for such noisy data!

@FunAILab with @valeoai with first author @FragileGoodwill and help from @IdeKohsuke @gatheluck @HirokatuKataoka Gilles Puy @abursuc

12hViews 1.5KLikes 37Bookmarks 3