Temporal Object-Aware Vision Transformer for Few-Shot Video Object Detection

Yogesh Kumar, Anand Mishra · 2025

Read Temporal Object-Aware Vision Transformer for Few-Shot Video Object Detection by Yogesh Kumar, Anand Mishra (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-2511-13784.txt.

Description

Yogesh Kumar and Anand Mishra present a temporal object-aware vision transformer for detecting new object types in video from only a few labeled examples. Few-shot video detection must hold identity steady across frames despite occlusion and changing looks, without pricey region proposals. Their filter passes only high-confidence object features forward through time, which limits noise build-up.

Detection and classification heads trained in the few-shot regime read those propagated features, giving steady tracks without explicit tube proposals. Gains appeared across four benchmarks in the five-shot setting, with further wins at one, three, and ten shots. The code is public, inviting reuse for low-label video tasks in surveillance and media search.

Catalog record

CREATORYogesh Kumar, Anand Mishra
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-13784

Browse the full library · About the Tower of Babel