Monocular Metric Depth Still Fails on 7.7% of Everyday Scenes, Mostly Scale Errors
ducha_aiki · x · 2026-11-19
An arXiv paper, "What the Averages Hide," argues metric depth estimation is not solved in everyday scenes — the remaining difficulty is mostly scale.
- Proposes per-image evaluation: failure rate (metric δ₁ < 0.5), scale-aligned structure metrics, and cross-model comparison separating shared from model-specific failures
- Evaluates 11 recent architectures (18 evaluations, with/without GT intrinsics) on 7 everyday-scene benchmarks (4,406 images) used by MoGe-2
- The best evaluation still fails 7.7% of images; 78.8% of failures have intact structure, and 98.3% vanish after scale correction — errors are almost purely scale
- Images failed by ≥6 architectures make up 7.3% of data (concentrated on HAMMER and DIODE), carrying 34–40% of top-6 models' error even when selected by other families
- Adding objects to a scene helps scale estimation
More from Research
- RhoAgent in Nature Biotech brings AI design from proteins to programmable functional RNA — EricTopol · 2026-11-19
- Designing a humanoid robot foot: heel strikes hit 1511N, whole foot weighs 117g — dee_hw · 2026-11-19
- Google's GUARD paper: frozen base model, self-editing agent harness gains +8.89 points — omarsar0 · 2026-11-19
- Kezek splits mono two-person conversations into per-speaker tracks for full-duplex dialogue training data — manifestro · 2026-11-19
- Elie Bakouch: current AI research benchmarks overfit and can't track RSI — eliebakouch · 2026-11-19
- Hutter Prize text compression jumps ~10% by overfitting small transformers for arithmetic coding — ricklamers · 2026-11-19