Conference on Computer Vision and Pattern Recognition (CVPR)

What it is

CVPR, the Conference on Computer Vision and Pattern Recognition, is an annual conference on computer vision and pattern recognition. It is normally held in June in North America^[1]. Computer vision is the field that tries to get machines to extract information from images and video; pattern recognition is about finding regularities in data and using them to classify or decide. The two areas share methods and test sets, so a single venue covers both.

Origins

The conference was first held in 1983 in Washington, DC. It was organized by Takeo Kanade and Dana H. Ballard^[1].

Who runs it

From 1985 to 2010 the conference was sponsored by the IEEE Computer Society^[1]. In 2011 it was also co-sponsored by the University of Colorado Colorado Springs^[1]. Since 2012 it has been co-sponsored by the IEEE Computer Society and the Computer Vision Foundation, which provides open access to the conference papers^[1]. Open access means the papers can be read and downloaded by anyone, rather than sitting behind a subscription that only well-funded labs can afford.

The conference is managed by a rotating group of volunteers, chosen in a public election at the PAMI-TC meeting four years before the meeting they will run^[1]. Choosing the team four years ahead gives it time to plan and to recruit reviewers.

Review is multi-tier and double-blind^[1]. Double-blind means reviewers do not know who wrote a paper, and authors do not know who reviewed it, so judgement rests on the work rather than on the reputation of a known group. Multi-tier means a paper's reviews are collected and weighed at more than one level before a decision is reached, rather than resting on a single reader.

Acceptance is selective: the conference generally has less than 30% acceptance rates for all papers, and less than 5% for oral presentations^[1]. A paper given an oral slot is presented to the whole audience; the rest of the accepted papers are still shown, but in a less central format.

Awards

Its awards include the Longuet-Higgins Prize, given to papers from ten years earlier that made a significant impact; the PAMI Young Researcher Award; and the PAMI Thomas S. Huang Memorial Prize, which was established at the 2020 conference^[1]. The ten-year look-back is a way of rewarding work whose importance has held up over time rather than only at the moment of publication, while the other two recognise researchers earlier in their careers and remember a member of the community.

One edition: CVPR 2016

CVPR 2016 was the 29th IEEE Conference on Computer Vision and Pattern Recognition, held in Las Vegas from June 26 to July 1, 2016^[2]. Alongside its main four-day program it included 29 workshops and 22 tutorials^[2]. Workshops are smaller meetings on narrower topics, where a specialised group can compare notes; tutorials are teaching sessions that introduce the background and tools of an area, which is useful for students and for people moving into it from neighbouring fields.

Place in the field

The numbers describe a venue that accepts a minority of what it receives, and gives oral slots to a small fraction of that minority^[1]. Because the Computer Vision Foundation provides open access to the papers, the results presented there are reachable by readers outside the institutions that can pay for subscriptions, which matters for students and for people in industry who want to check a method for themselves^[1]. The management model — volunteers elected years in advance — is one way of running a large annual meeting without a permanent staff. The rotating team also means the conference is not owned by any single group of researchers, and the double-blind process is intended to keep the selection focused on the submitted work^[1].