AI Paper Atlas

Research map / 3D vision / Depth and stereo

DA$^{2}$: Depth Anything in Any Direction

Haodong Li, Wangguandong Zheng, Jing He, Yuhao LIU, Xin Lin, Xin Yang, Ying-Cong Chen, Chunchao Guo

ICLR 2026 · 3D vision › Depth and stereo · approach: Depth from light field and defocus

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
convert perspective depth datasets to panoramas and add to training (63K -> 607K samples) (ablation)DA2 · Average of Stanford2D3D, Matterport3D, PanoSUNCGAbsRel (%): 8.07 -> 6.62Improved
add Hypersim converted to panorama to Structured3D (ablation)DA2 · Average of 3 benchmarksAbsRel (%): 8.07 -> 7.10Improved
add Dynamic Replica panoramas (largest curated set) after 316,722 samples (ablation)DA2 · Average of 3 benchmarksAbsRel (%): 6.66 -> 6.62No clear effect
add panoramic out-painting (FLUX-I2P) of RGB images in curation (ablation)DA2 · Average of 3 benchmarksAbsRel (%): 7.59 -> 6.62Improved
add spherical embedding via cross-attention in SphereViT (ablation)DA2 · Average of 3 benchmarksAbsRel (%): 6.84 -> 6.62Improved
add auxiliary normal loss (ablation)DA2 · Average of 3 benchmarksAbsRel (%): 6.99 -> 6.62Improved
add spherical embedding via cross-attention in SphereViT (ablation)DA2 · Average of 3 benchmarksRMSE: 20.87 -> 20.63Improved
use P2E projection only for GT depth, no out-painting of depthData curation engine · Data curation designdepth GT accuracyNo clear effect
train end-to-end zero-shot DA2 vs strongest zero-shot baselineDA2 · Zero-shot panoramic benchmarksAbsRel: average 38% improvementImproved
use L1 normal loss instead of angular lossDA2 · Training loss designtraining stabilityImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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