Monocular Normal Estimation via Shading Sequence Estimation
Zongrui Li 1 Xinhua Ma 1 Minghui Hu 1 Yunqing Zhao 2 Yingchen Yu 2 Qian Zheng 3,4 Chang Liu 5 Xudong Jiang 1 Song Bai 2 1 School of Electrical and Electronic Engineering, Nanyang Technological University 2 ByteDance 3 College of Computer Science and Technology, Zhejiang University 4 The State Key La · 2026
Read Monocular Normal Estimation via Shading Sequence Estimation by Zongrui Li 1 Xinhua Ma 1 Minghui Hu 1 Yunqing Zhao 2 Yingchen Yu 2 Qian Zheng 3,4 Chang Liu 5 Xudong Jiang 1 Song Bai 2 1 School of Electrical and Electronic Engineering, Nanyang Technological University 2 ByteDance 3 College of Computer Science and Technology, Zhejiang University 4 The State Key La (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-09929.txt.
Description
Estimating a surface's normals from a single photo is a staple of computer vision, yet current models often get the look right while the rebuilt shape drifts. A 2026 paper led by Zongrui Li, with eight co-authors, traces the fault to the task itself. Geometry shows up only as faint color shifts, so a model struggles to separate one shape from another. The team reframes the problem around shading instead.
Their method, RoSE, uses image-to-video generative models to predict a sequence of shadings under varied light. It then converts that sequence into a normal map with a simple least-squares solve. Training on a synthetic set called MultiShade, with varied shapes, materials, and lighting, hardens it for complex objects. Tests report leading results on real-world benchmarks for object-based normal estimation.
Catalog record
| CREATOR | Zongrui Li 1 Xinhua Ma 1 Minghui Hu 1 Yunqing Zhao 2 Yingchen Yu 2 Qian Zheng 3,4 Chang Liu 5 Xudong Jiang 1 Song Bai 2 1 School of Electrical and Electronic Engineering, Nanyang Technological University 2 ByteDance 3 College of Computer Science and Technology, Zhejiang University 4 The State Key La |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-09929