arxivcs.CV2026-07-06
Video-Text Temporal Localization via Multi-Scale Convolution and Dynamic Routing
Gengtian Shi, Jinze Yu, Chenhao Wu, Shaofei Wang, Eiji Fukuzawa, Junjie Tang, et al.
Video-text temporal localization requires precise alignment between natural language queries and corresponding video segments, a fundamental challenge in multimodal understanding. We present a novel framework that addresses two critical limitations of existing methods: inadequate…