SHADeS: Self-supervised Monocular Depth Estimation Through Non-Lambertian Image Decomposition

Rema Daher, Francisco Vasconcelos, Danail Stoyanov · 2025

Read SHADeS: Self-supervised Monocular Depth Estimation Through Non-Lambertian Image Decomposition by Rema Daher, Francisco Vasconcelos, Danail Stoyanov (2025) 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-2502-12994.txt.

Description

Rema Daher, Francisco Vasconcelos, and Danail Stoyanov build a self-teaching model to map the inside of the colon in 3D from single camera images, in a 2025 paper. Such maps could help doctors see which parts of the colon a scope has covered and judge polyp size and shape. The colon’s wet walls throw bright reflections that fool depth tools.

Their model splits each image into shading, surface color, depth, and shiny highlights, learning all four at once with no hand labels. Unlike older splits, theirs treats the shine as its own light part rather than smearing it into color. Tests on real colonoscopy video show the split stays stable under harsh light changes. Cleaner depth maps are the payoff for navigation aids.

Catalog record

CREATORRema Daher, Francisco Vasconcelos, Danail Stoyanov
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-12994

Browse the full library · About the Tower of Babel