UAMP: Consistent video object segmentation with uncertainty-aware memory propagation

Yichuang Luo, Fang Wang, Xiaohu Liu · 2026

Read UAMP: Consistent video object segmentation with uncertainty-aware memory propagation by Yichuang Luo, Fang Wang, Xiaohu Liu (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/plos-pone-0353156.txt.

Description

Yichuang Luo, Fang Wang, and Xiaohu Liu present this 2026 PLOS One computer vision paper on following objects through video reliably, even when they are hidden briefly or move fast. Their method upgrades a leading segmentation model whose memory of earlier frames can accumulate errors. The upgrade estimates uncertainty about each object's appearance and motion and uses it to decide what the model should remember, with separate long-term and short-term memory handling.

Across five standard video benchmarks the method beat previous state-of-the-art systems, with the largest gains in scenes involving occlusion and objects reappearing after vanishing. The authors argue that tracking what a model is unsure about, not just what it sees, is the key to segmenting crowded, fast-changing video.

Catalog record

CREATORYichuang Luo, Fang Wang, Xiaohu Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE60567
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0353156

Browse the full library · About the Tower of Babel