HTTP/2 200 server: GitHub.com content-type: text/html; charset=utf-8 last-modified: Sat, 27 Sep 2025 17:55:45 GMT access-control-allow-origin: * strict-transport-security: max-age=31556952 etag: W/"68d82521-615b" expires: Sun, 28 Dec 2025 23:27:03 GMT cache-control: max-age=600 content-encoding: gzip x-proxy-cache: MISS x-github-request-id: 3E6B:318CF6:802481:8FFE7B:6951BA6B accept-ranges: bytes age: 0 date: Sun, 28 Dec 2025 23:17:03 GMT via: 1.1 varnish x-served-by: cache-bom-vanm7210097-BOM x-cache: MISS x-cache-hits: 0 x-timer: S1766963823.386330,VS0,VE204 vary: Accept-Encoding x-fastly-request-id: 5a9282e83eaf3033cd76e69b64543cb643293761 content-length: 4046

Easi3R: Estimating Disentangled Motion from DUSt3R Without Training

Xingyu Chen¹     Yue Chen¹     Yuliang Xiu^1,2     Andreas Geiger³     Anpei Chen^1,3    

¹Westlake University      ²Max Planck Institute for Intelligent Systems      ³University of Tübingen, Tübingen AI Center

arXiv Project Page ⚖️ Comparison Code

Disentangled Reconstruction

We introduce the spatial and temporal attention mechanism behind DUSt3R, enabling robust dynamic object segmentation.
This segmentation efficiently disentangles object and camera motion, enabling DUSt3R to robustly reconstruct 4D.

4D Reconstruction

Static Scene

Dynamic Object

Results are downsampled for efficient online rendering

Left-click and drag to rotate

Right-click and drag or WASD to move

Scroll to zoom

Click to pause