This repo hold the code, the models and the evaluation kit for the MUSES (MUlti-Shot EventS) dataset, introduced in the paper:
Multi-shot Temporal Event Localization: a Benchmark
Xiaolong Liu, Yao Hu, Song Bai, Fei Ding, Xiang Bai, Philip H.S. Torr
MUSES is a large-scale video dataset, designed to spur researches on a new task called multi-shot temporal event localization. Refer to the paper and the project page for more information.
The code will be released in May 2021.