<!DOCTYPE html>
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
</head>
<body>
[Apologies if you receive multiple copies]<br>
<br>
<b>Call for Papers - ILR+G@ICCV2025</b><br>
<br>
<b>7th Instance-Level Recognition and Generation Workshop</b><br>
International Conference on Computer Vision, <a
href="https://iccv.thecvf.com/">ICCV 2025</a><br>
Honolulu, Hawaii, October 19-20 2025<br>
<a class="moz-txt-link-freetext" href="https://ilr-workshop.github.io/ICCVW2025/">https://ilr-workshop.github.io/ICCVW2025/</a><br>
<br>
<b>Important dates</b><br>
<ul>
<li>submission deadline: June 21, 2025</li>
<li>notification of acceptance: July 10, 2025</li>
<li>camera-ready papers due: August 17, 2025</li>
</ul>
<br>
<b>Topics</b><br>
The Instance-Level Recognition and Generation (ILR+G) Workshop aims
to explore computer vision tasks focusing on specific instances
rather than broad categories, covering both recognition
(instance-level recognition - ILR) and generation (instance-level
generation - ILG). Unlike category-level recognition, which assigns
broad class labels based on semantics (e.g., “a painting”), ILR
focuses on distinguishing specific instances, assigning class labels
that refer to particular objects or events (e.g., “Blue Poles” by
Jackson Pollock), enabling recognition, retrieval, and tracking at
the finest granularity. This year, the workshop also covers ILG,
also known as personalized generation, which involves synthesizing
new media that preserve the visual identity of a particular instance
while varying context or appearance, often guided by text. We
encourage the exploration of synergies between ILR and ILG, such as
using recognition as a foundation for instance-conditioned
generation, or leveraging generative models to boost ILR in low-data
or open-set scenarios.<br>
<br>
Relevant topics include (but are not limited to):<br>
<ul>
<li>instance-level object classification, detection, segmentation,
and pose estimation</li>
<li>particular object (instance-level) and event retrieval</li>
<li>personalized (instance-level) image and video generation</li>
<li>cross-modal/multi-modal recognition at instance-level</li>
<li>other ILR tasks such as image matching, place recognition,
video tracking, moment retrieval</li>
<li>other ILR+G applications or challenges</li>
<li>ILR+G datasets and benchmarking</li>
</ul>
The task of person and vehicle re-identification clearly falls
within our definition of ILR. Nevertheless, because of its social
implications, we intentionally omit it from the list of topics.<br>
<br>
<b>Submission details</b><br>
We call for novel and unpublished work in the format of long papers
(up to 8 pages) and short papers (up to 4 pages). Papers should
follow the <a
href="https://iccv.thecvf.com/Conferences/2025/AuthorGuidelines">ICCV
proceedings style</a> and will be reviewed in a double-blind
fashion. Selected long papers will be invited for oral
presentations; all accepted papers will be presented as posters.
Only long papers will be published in the ICCV workshop proceedings.
All submissions will be handled electronically via the <a
href="https://cmt3.research.microsoft.com/ILRnG2025">CMT
conference submission website</a>.<br>
<b><br>
Organizing committee</b><br>
Andre Araujo, Google DeepMind<br>
Bingyi Cao, Google DeepMind<br>
Kaifeng Chen, Google DeepMind<br>
Ondrej Chum, Czech Technical University in Prague<br>
Noa Garcia, Osaka University<br>
Guangxing Han, Google DeepMind<br>
Giorgos Kordopatis-Zilos, Czech Technical University in Prague<br>
Giorgos Tolias, Czech Technical University in Prague<br>
Hao Yang, Amazon<br>
Nikolaos-Antonios Ypsilantis, Czech Technical University in Prague<br>
Xu Zhang, Amazon<br>
<br>
</body>
</html>