AI Paper Atlas

Research map / 3D vision / Gaussian splatting

Video2Robo: 3DGS-based Synthetic Data from One Video Enables Scalable Robot Learning

Yinan Deng, Kejia Hu, Ye Chen, Jianyu Dou, Jiahui Wang, Jingyu Zhao, Haojia Ao, Yi Yang, Yufeng Yue

CVPR 2026 · 3D vision › Gaussian splatting · approach: Monocular dynamic scene reconstruction

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use Video2Robo 3DGS tracking with CoTracker3 track loss instead of FoundationPose-O6D object localization · 6D localization, mean of 6 tasksBOP Average Recall: 55.36 -> 78.41Improved
use Video2Robo tracking instead of MegaPose6D object localization · 6D localization, StackBOP Average Recall: 91.57 -> 90.90No clear effect
use 3DGS rendering and kinematic motion modeling instead of simulator-based MimicGendata generation · 6 self-captured tasksaverage time per sample (s): 17.23 -> 05.23Improved
use robot-object coupled motion instead of simulated physics for data generationdata generation · 6 self-captured tasksgeneration success rate (%): 97.40 -> 100.00Improved
add texture, background, lighting and camera augmentation to MimicGen (ablation)MimicGen · Simulation, various changessuccess rate (%): 11.67 -> 29.67Improved
add appearance and view augmentation to MimicGen (ablation)MimicGen · Simulation, only object poses changesuccess rate (%): 93.00 -> 25.67Hurt
train on Video2Robo data rather than MimicGen/SkillGen dataDiffusion Policy · Simulation, various changessuccess rate (%): 44.67 -> 63.83Improved
train on Video2Robo data rather than MimicGen/SkillGen dataDiffusion Policy · Simulation, only object poses changesuccess rate (%): 93.00 -> 66.67Hurt
train on Video2Robo data instead of 20 real teleoperated demosDiffusion Policy · Real world, various changessuccess rate (%): 27.50 -> 58.33Improved
train on Video2Robo data instead of simulation baselinesDiffusion Policy · Real world, various changessuccess rate (%): 5.83 -> 58.33Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy