Computer Vision News - November 2021

2 Summary Poster Presentation 32 Best of ICCV 2021 Film editors are presented with several shots taken from different cameras and have to decide where to cut in each of these shots to then put them all together to form a scene. This process is very long and requires a lot of manual work. Currently there are software tools that help, but there is no automated process. A human has to do everything. Given a pair of shots from two different cameras, Alejandro and his team propose the first automated solution to this problem . To do this, the team not only need movies, but they need all the raw recordings from the movies as well. Although there are lots of movie clips available on the internet, this raw footage is harder to find . Instead, they download scenes from YouTube and use a shot detector, which tells you when there is a camera change in a scene. Learning To Cut by Watching Movies Alejandro Pardo is a third-year PhD student at King Abdullah University of Science and Technology (KAUST) in Saudi Arabia, under the supervision of Professor Bernard Ghanem. His work is the first to tackle an important video editing problem called ranking cut plausibility. It has been accepted as a poster and he spoke to us ahead of his live Q&A session. Movies often record 10 times more footage than we see in the final cut. For a documentary, it can be 100 times more. So, for every hour we watch, 100 hours of footage will have been recorded . This is a big problem for editors.