Reproduce “EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jiawei Yang, Boris Ivanovic, Or Litany, Xinshuo Weng, Seung Wook Kim, Boyi Li, Tong Che, Danfei Xu and 3 more
EmerNeRF learns a time-aware representation of driving scenes that separates persistent structure from moving content. It predicts 3D scene flow and uses it to aggregate dynamic features across nearby moments, without ground-truth masks or flow labels. On a shorter-training ablation, flow accuracy reaches 89.74%, versus 4.53% without aggregation and 3.87% with gradients stopped; this supports the mechanism in that experiment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.