Video summarization based on multi-scale feature fusion

Jing Bao, Shipeng Xu, Jing Zhang · 2026

Read Video summarization based on multi-scale feature fusion by Jing Bao, Shipeng Xu, Jing Zhang (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-0353773.txt.

Description

Bao and colleagues, in a 2026 paper, build a model that summarizes videos by picking their most important shots. A good summary lets a viewer grasp a long video in moments, which matters for surveillance review and short-video apps. Earlier graph-based methods struggled to blend close-up detail with the whole video's shape.

Their model turns each shot into a point in a graph, linked to shots with similar meaning. It then reads the graph at two scales: one attention method draws out local links between neighboring shots, and another network draws out global structure. The three views, original, local, and global, are fused to score each shot's importance. On two public benchmark sets, the method scored competitively against leading approaches. The authors offer it as a practical way to cut long footage down to its essentials.

Catalog record

CREATORJing Bao, Shipeng Xu, Jing Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45475
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0353773

Browse the full library · About the Tower of Babel