Motions: CVPR 2024 PAMI-TC Meeting

Call for Motions: CVPR 2024 PAMI-TC Meeting

This year's PAMI-TC meeting is scheduled for 4-5pm on Thursday, June
20th (Summit Flex Hall ABC). Please mark this on your calendars and
plan to attend if you are interested in the operations and policies of
the computer vision community. An attempt will be made simultaneously
broadcast the meeting live on the virtual conference platform for
remote attendees.

All attendees of CVPR 2024 are eligible to vote for the motions
specific to the June TC meeting listed below, which will be discussed
at the meeting.


Motion 1: Justification of Closed-Source Comparisons

Proposed by Andrea Tagliasacchi (Simon Fraser University, University
of Toronto), Torsten Sattler (Czech Technical University in Prague),
Marcus Brubaker (York University), Dmytro Mishkin (Czech Technical
University in Prague), Kosta Derpanis (York University), Marc
Pollefeys (ETH Zurich), Yasutaka Furukawa (Simon Fraser University),
Konrad Shindler (ETH Zurich), Andreas Geiger (University of Tubingen),
Kwang Moo Yi (University of British Columbia), David Lindell
(University of Toronto), Or Litany (Technion), Eric Brachmann
(Niantic), David Fleet (University of Toronto), Siyu Tang (ETH Zurich)

The PAMI TC recommends to the ICCV / CVPR Steering Committee:
Effective with the CVPR 2025 conference, whenever a comparison of
published research without publicly available code / data / pretrained
models is requested (i.e., requiring re-implementation), it should be
appropriately justified if used as a basis for a paper decision. This
is unless this is a somewhat minor change to an already implemented
method for which the code / data is available or re-implementing a
method based on details provided by a publication is common practice
in a sub-field. In any case, comparisons should only be requested if
the publication and / or code has been available sufficiently ahead of
the submission deadline.

Background: It is becoming increasingly common that reviewers mention
a quantitative comparison against a paper as a reason for rejection,
even if such paper does not have published code, does not provide a
pre-trained model to evaluate against, was trained on a private
dataset, and / or does not provide sufficient details to replicate the
method or the experimental settings. Even worse, reviewers have
requested comparisons to closed-source software that does not even
have a white paper or scientific publication.

The current pace of development in Computer Vision and Pattern
Recognition is unprecedented, and much of this has been made possible
thanks to open-source code. We want to further promote openness,
without mandating every submission to provide a public implementation.

More detailed explanation and implementation details can be found
here: https://tinyurl.com/2648k253.


Motion 2: Text-Only Rebuttals

Proposed by Ramin Zabih (Cornell University)

The PAMI TC recommends to the ICCV / CVPR Steering Committee: All
future editions of CVPR and ICCV should limit the rebuttal to text
only, following best practices from SIGGRAPH, in order to better
prevent reviewers from requesting additional experiments in the
rebuttal.

Motivation: The PAMI TC passed a motion in 2018 that was intended to
forbid additional experiments during the rebuttal phase of the review
process and ensure that reviewers do not penalize papers due to the
lack of additional experiments. However, such requests have persisted
in subsequent conference review cycles, with reviewers, area chairs,
and program chairs all accepting new experiments when they should not
be. In order to further reduce the ability of reviewers to make
requests for additional experiments in the rebuttal phase, this motion
institutes a SIGGRAPH style text-only rebuttal that would eliminate
visual data.


Motion 3: The Location of ICCV 2029

Proposed by Gerard Medioni (Chair of the ICCV/CVPR Steering Committee)

The ICCV / CVPR Steering Committee made a preliminary decision to hold
ICCV 2029 in Dubai, with ICCV 2031 to be held in Europe. Following the
discussion held during the PAMI TC meeting, please vote whether or not
to recommend that the Steering Committee move forward with the Dubai
venue for ICCV 2029.

Motivation: For the first time since its inception, the CVPR / ICCV
Steering Committee selected a location for an upcoming edition of
ICCV. As with the rollout of any new organizational structure, there
were bound to be some hiccups. And that is what happened with this
first attempt at selecting a venue.

The steering committee voted to approve an ICCV 2029 proposal with a
strong team of GCs and PCs and Dubai as the location. The location is
attractive for several reasons, including an easy visa process for
many countries, many flights, and high quality of conference venue. To
date, no contract has been signed for the venue. After the
announcement of the location, concerns were raised to the committee
about this particular location in regard to the community's diversity
and inclusion goals, as well as the lack of transparency in the
selection process.

In response, the steering committee is raising this motion to gauge
community sentiment about the location. In the event that the steering
committee withdraws its approval of the Dubai location, it will
quickly find at least one new location for ICCV 2029 (with an option
to keep the current ICCV 2029 organizing team, should the team be
willing to pivot to a new venue), which will be circulated to the
community via the PAMI TC newsletter before the steering committee
approves it. If the motion is approved, the steering committee intends
to hold ICCV 2031 in Europe.