Computer Vision News - August 2017

44 / 56

Every month, Computer Vision News reviews a challenge related to our field. If
you do not take part in challenges, but are interested to know the new methods
proposed by the scientific community to solve them, this section is for you. This
month we dedicate this space to the DAVIS workshop and challenge, organized
as a satellite event of CVPR2017. We published a preview of the event and
now Jordi Pont-Tuset tells you what actually happened at the workshop.
The 2017 DAVIS Challenge on Video Object Segmentation workshop
) was held on the morning of the last day of CVPR2017. Video
object segmentation is about separating the pixels of an object from the
background in a video, and in particular, the challenge is about the semi-
supervised approach, where the segmentation of the first frame is given and it
should be propagated to the rest of the video.
The DAVIS dataset consists of 150 high-resolution video sequences, annotated
with pixel-level accurate segmentations, totalling more than 10.000 annotated
frames. The participants were given a subset of the annotations and they had
the submit the results for the rest of the subset, for which the annotations are
not public. The result could be submitted to an evaluation server that returned
the quality of the result.
Challenge
Challenge - DAVIS
44
Densely Annotated Video Object Segmentation
by Jordi Pont-Tuset
BEST OF CVPR