The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) is regarded as the leading conference in this field. In common with many other industrial research organizations such as Google, Facebook, IBM or Nvidia, Konica Minolta Lab Europe took part in this event, and I had the opportunity to present our activities on Automatic guidance of workers using Augmented Reality in one of the side workshops of the conference.
The 2017 CVPR conference, held from 22nd to 25th July in Honolulu, Hawaii, has been huge: the largest CVPR ever. To mention some figures:
- More than 5,000 people, an increase of more than 33% compared to the last year.
- 783 papers presented during the main conference (22% more than last year), 71 of these as oral presentation, other 144 as spotlight presentations.
- Each of the 783 papers had a poster, with three parallel sessions during the first two days and two parallel sessions during the last two days: a solution that enabled one not to miss any paper even in the event of overlapping presentations at the same time.
- 44 workshops and 20 tutorials before and after the conference.
This was a dense and exciting programme for anyone interested in the Computer Vision (CV) area. Moreover, several companies presented their technologies and solutions within a variety of applications during the Industry EXPO, thus demonstrating the increasing interest of industry in the Computer Vision domain. Another piece of information confirming that CV has definitely entered the business domain is that the amount of CVPR funding derived from sponsors has increased by 79% with respect to last year to reach USD 859,000.
— KM Labs Europe (@KMLabsEU) August 4, 2017
The most common application for many exhibitors was, perhaps not surprisingly, autonomous driving. Both small and large organisations are dedicating great efforts towards this application and I am really looking forward to seeing how this area will evolve in the next year.
Most of the topics for the presentations in the overall conference centred on object recognition and scene understanding, two contexts that have recently benefitted from the successful application of Deep Learning methods. Beyond these areas of focus, I have to admit that Deep Learning algorithms have been discussed and presented everywhere in CVPR, from 3D-vision to low-level vision, for robotic vision, for object tracking, and for computational photography, and many other areas.