4D Reconstruction and Tracking
Participants reconstruct dynamic scenes and track points over time from an input video, following the D4RT-style reconstruction and tracking setting.
A synthetic benchmark with high-quality multi-view rendering and dense geometric annotations, designed for reconstructing the physical world in motion.
Both tracks use the PhysAI dynamic benchmark and share a common goal: robust physical-world understanding from video.
Participants reconstruct dynamic scenes and track points over time from an input video, following the D4RT-style reconstruction and tracking setting.
Models generate a new-view video from an input video and a target camera trajectory, following the camera-controlled dynamic rendering setting.
Please also submit a technical report to OpenReview describing your methodology before the submission deadline. Kindly use the provided template and ensure that your team name is clearly included in the report.
Teams that demonstrate strong performance and submit high-quality technical reports will be invited to contribute to the challenge's final technical report. All authors from selected teams will be included as co-authors of the final report.
All deadlines are 23:59 UTC (11:59 PM UTC).
For questions about the benchmark, participation, or the technical report, please reach out: