AI Paper Atlas

Research map / Video / Video segmentation and tracking

PVDepth: Panoramic Video Depth Estimation via Geometry-Aware Spatiotemporal Adaptation

Chuanxin Song, Peixi Peng

ICML 2026 · Video › Video segmentation and tracking · approach: Layered video decomposition and dynamics

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
add PSNI noise initialization (ablation)DepthCrafter baseline · PanoCARLA (avg of 3 settings)AbsRel: 0.248 -> 0.228Improved
add CRTM cube temporal branch (ablation)DepthCrafter baseline · PanoCARLA (avg of 3 settings)AbsRel: 0.248 -> 0.208Improved
add CRTM on top of PSNI (ablation)PVDepth · PanoCARLA (avg of 3 settings)AbsRel: 0.228 -> 0.199Improved
extend baseline training from 20k to 40k steps (ablation)DepthCrafter baseline · PanoCARLA (avg of 3 settings)AbsRel: 0.248 -> 0.232Improved
anneal spherical noise instead of using it directly (ablation)PSNI · PanoCARLA (avg of 3 settings)AbsRel: 0.235 -> 0.228Improved
add residual normal noise to annealed noise (ablation)PSNI + CRTM · PanoCARLA (avg of 3 settings)AbsRel: 0.203 -> 0.199Improved
use projection layer rather than naive average fusion (ablation)CRTM · PanoCARLA (avg of 3 settings)AbsRel: 0.267 -> 0.214Improved
use projection layer rather than naive add fusion (ablation)CRTM · PanoCARLA (avg of 3 settings)AbsRel: 0.356 -> 0.214Improved
use cube features as key/value source instead of equi (ablation)CRTM · PanoCARLA (avg of 3 settings)AbsRel: 0.214 -> 0.208Improved
two-stage training instead of one-stage joint (ablation)PVDepth · PanoCARLA (avg of 3 settings)AbsRel: 0.221 -> 0.199Improved
freeze spatial layers in stage 2 (ablation)PVDepth · PanoCARLA (avg of 3 settings)AbsRel: 0.215 -> 0.199Improved
add PSNI under two-stage frozen protocol (ablation)PVDepth · PanoCARLA (avg of 3 settings)AbsRel: 0.204 -> 0.199Improved

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