Welcome to CVPR 2012 in Providence, Rhode Island. We received 1933 completed submissions to the conference, a substantial increase over the previous year. To select papers from these submissions, we invited 45 well-known vision researchers to act as Areas Chairs (ACs) and recruited an expert team of 935 reviewers from the broader computer vision community. Recognizing the crucial importance of qualified reviewers to the review and decision process, the initially compiled reviewer pool was first vetted by the Program Chairs through cross-checking a reviewer's recent publications in a number of major computer vision related conferences and journals, and then augmented by additional reviewers recommended by the ACs. We again used the CMT conference management service sponsored by Microsoft Research to manage the submission and selection of papers from beginning to end.
After the submission deadline, the papers were distributed between the ACs for the first time using an automated assignment system, the Toronto Paper Matching System. This system, based on the optimization method of Charlin et al. UAI 2011,suggests matches between papers and reviewers (ACs, in our case) based on bag-of-words descriptors extracted from the PDF files of submitted manuscripts and representative publications by each potential reviewer. The ACs in turn identified potential (and non-conflicted) reviewers for each of their assigned papers, from which the CMT system automatically selected three reviewers per paper. Extensive manual adjustments were made by the ACs and Program Chairs to achieve better matches between the papers and reviewers under the workload constraints. Reviewers were given seven weeks to complete their reviews, at which time the ACs stepped back in to finish their work: consolidating reviews and author rebuttals, initiating discussions for clarification, and making recommendations for decisions on papers. The Program Chairs and the ACs worked extremely hard to ensure that every paper eligible for full review received at least three reviews.
The decision process was designed to ensure that every paper and its reviews and author rebuttal were looked at by at least two ACs. To further support a thorough review process, at the AC Meeting in Long Beach, CA, we divided the ACs into four panels, with almost no conflict between the ACs and papers associated with each panel. The Program Chairs served as the panel chairs and worked hard to maintain consistency between the panels. All decisions were made by at least two ACs working together and, as needed, by the whole panel. A consensus of the entire panel was sought on the most difficult cases. The ACs produced detailed consolidation reports to justify all decisions. The Program Chairs and General Chairs did not submit any papers to CVPR 2012, allowing them to work without any direct conflicts throughout the review process. Additionally, the respective panel chairs were excluded from any decisions associated with papers from their affiliated institutions. The double-blind nature of the CVPR review process was strictly maintained throughout the review process.
The ACs accepted 48 papers as orals (2.5%) and 418 papers as posters (21.6%), with an overall acceptance rate of 24.1%. There was no prior constraint, either on the number of orals or posters.
We are publishing the proceedings in USB drive form. All published papers in the main conference and associated workshops will be indexed by the IEEE, and available through the IEEE Digital Library. We wish to thank the other members of the Organizing Committee, the Area Chairs, Reviewers, Authors, and the CMT team for the immense amount of hard work and professionalism that has gone in to making CVPR 2012. Our thanks also go to the organizers of previous CVPRs for their helpful advice and support. Finally, we wish all the delegates a highly stimulating, informative, and enjoyable conference.