Remote SAMsing: From Segment Anything to Segment Everything Journal: International Journal of Applied Earth Observation and Geoinformation

Osmar Luiz Ferreira de Carvalho, Osmar Abílio de Carvalho Júnior, Anesmar Olino de Albuquerque, Daniel Guerreiro e Silva · 2026

Read Remote SAMsing: From Segment Anything to Segment Everything Journal: International Journal of Applied Earth Observation and Geoinformation by Osmar Luiz Ferreira de Carvalho, Osmar Abílio de Carvalho Júnior, Anesmar Olino de Albuquerque, Daniel Guerreiro e Silva (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-2605-00256.txt.

Description

Osmar Luiz Ferreira de Carvalho and 3 co-authors adapt a strong image segmentation model, SAM2, to the huge scenes captured by satellites and aircraft. Two problems arise: strict settings give precise masks but leave most of a scene unmarked, while loose settings cover more at lower quality. Tiling the scene also splits objects across tile edges.

Their open-source pipeline, Remote SAMsing, runs the model over each tile again and again, painting accepted masks black so later passes see a simpler scene. Thresholds loosen only when coverage stalls, so the sharpest masks come first. Padding and a matching step rebuild objects split by tiles. On seven scenes, coverage rose from well under two-thirds to above nine-tenths in most cases.

Catalog record

CREATOROsmar Luiz Ferreira de Carvalho, Osmar Abílio de Carvalho Júnior, Anesmar Olino de Albuquerque, Daniel Guerreiro e Silva
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.00256

Browse the full library · About the Tower of Babel