Temporal and Contextual Transformer for
Multi-Camera Editing of TV Shows
ECCVW 2022

Anyi Rao1     Xuekun Jiang2     Sichen Wang3     Yuwei Guo2     Zihao Liu3    
Dai Bo2     Long Pang3     Xiaoyu Wu3     Dahua Lin2,4 Libiao Jin3    



The ability to choose an appropriate camera view among multiple cameras plays a vital role in TV shows delivery. But it is hard to figure out the statistical pattern and apply intelligent processing due to the lack of high-quality training data. To solve this issue, we first collect a novel benchmark on this setting with four diverse scenarios including concerts, sports games, gala shows, and contests, where each scenario contains 6 synchronized tracks recorded by different cameras. It contains 88-hour raw videos that contribute to the 14-hour edited videos. Based on this benchmark, we further propose a new approach temporal and contextual transformer that utilizes clues from historical shots and other views to make shot transition decisions and predict which view to be used. Extensive experiments show that our method outperforms existing methods on the proposed multi-camera editing benchmark.

TVMCE Dataset


Overview of TV shows MultiCamera Editing (TVMCE) Datset. We reached out to film and TV production major colleges and follow their professional production team to acquire data covering various scenarios including concert, sports, contest and gala show that happen in universities and city theaters/stadiums. Our dataset holds a balanced coverage ratio among different scenario categories with 39% in gala shows and 14% in sports. Most shots in our dataset have a time duration between 0 to 8 seconds and a few shots are long shots that last longer than 32 seconds.



The website template was borrowed from Michaƫl Gharbi.