Hao Zhao (@HaoZhao_AIRSUN)
2025-12-08 | โค๏ธ 122 | ๐ 29
๐ DGGT: Feedforward 4D Reconstruction of Dynamic Driving Scenes using Unposed Images We introduce a fast, pose-free framework that reconstructs full 4D scenes in 0.4 s, enabling dynamic tracking and scene editing.
๐ State-of-the-art quality and efficiency on Waymo, nuScenes, Argoverse2.
๐ https://github.com/xiaomi-research/dggt
๐ ์๋ณธ ๋งํฌ
๋ฏธ๋์ด
