Overview
VAND brings together cutting-edge research on detecting what doesn’t belong in visual data—spanning anomaly, novelty, and out-of-distribution detection. Building on three successful editions, VAND 4.0 unites supervised, semi-, and unsupervised approaches, including few-, one-, and zero-shot learning, with a strong focus on real-world impact.
Join us for peer-reviewed papers, dataset-driven challenges, and invited talks on anomaly detection. This workshop brings together researchers from both industry and academia to present and discuss recent advances, opportunities, and open challenges in visual anomaly detection.
Invited Speakers
Walter Scheirer
Professor at
University of Notre Dame
Bodo Rosenhahn
Professor at
Leibniz University Hannover
Sebastian Höfer
Applied Science Manager at Amazon
David Zimmerer
PostDoc
Medical Image Computing Group, DKFZ
Talk Titles & Abstracts
Meaningful progress has been made in open world learning (OWL), enhancing the ability of agents to detect, characterize, and incrementally learn novelty in dynamic environments. However, novelty remains a persistent challenge for agents relying on state-of-the-art learning algorithms. This talk considers the current state of OWL, drawing on insights from a recent DARPA research program on this topic. I identify open issues that impede further advancements spanning theory, design, and evaluation. In particular, I emphasize the challenges posed by dynamic scenarios that are crucial to understand for ensuring the viability of agents designed for real-world environments. The talk provides suggestions for setting a new research agenda that effectively addresses these open issues.
Normalizing Flows belong to probabilistic machine learning models that convert a simple, sample-friendly distribution, such as a standard Gaussian, into a complex real-world data distribution through a series of invertible operations. Over the past years, we used such models for different anomaly detection settings and I will give an overview presentation on how our research emerged from a very simple baseline (called differnet from 2021) to more complex models (e.g. BUSSARD at cvpr-26). Several variants have been used in industry, they are fairly easy to train and use, and are fully open source.
Visual defect detection in retail logistics presents unique challenges: with millions of unique products — most observed only a handful of times — distinguishing "normal" from "defective" remains extremely difficult for AI systems. To advance research in this domain, we publicly released Kaputt, a large-scale dataset comprising 238K+ high-resolution images of 48K+ unique items, including 29K+ defective instances — 40× larger than previous benchmarks [Höfer et al. 2025]. The dataset captures real-world complexities ranging from minor creases to major spills. Kaputt also serves as the foundation for one of two tracks challenge part with this workshop, in the Visual Anomaly and Novelty Detection (VAND) 4.0 Challenge Retail Track. In this talk, we present the Kaputt dataset in detail, discuss the core challenges of defect detection in retail logistics, and highlight how VAND challenge participants have tackled the demanding tasks posed by this dataset.
Medical anomaly detection methods have long promised to revolutionize clinical workflows by supporting radiologists and highlighting overlooked findings. However, deploying these methods requires navigating unique domain-specific complexities, such as the needle-in-a-haystack nature of 3D volumes, detecting rare pathologies, and ensuring reliability in critical acute cases. This talk offers an overview of anomaly detection in radiological imaging and unpacks why current evaluation paradigms often fall short in this setting. I will present a framework for building benchmarks modeled after real-world clinical demands and map out the critical frontiers the computer vision community must conquer next.
Call for Papers
Our call for papers for VAND 2026 includes the following topics:
Anomaly detection, novelty detection, and out-of-distribution detection in images and videos.
Relevant learning paradigms including unsupervised, few-shot and active learning.
Dataset challenges including highly imbalanced data, noisy/incomplete labels, data sampling, applications spanning vision-based industrial inspection, predictive maintenance of complex machines.
Adjacent domains such as in-field inspection and medical diagnosis.
Theoretical contributions that address challenges unique to anomaly detection and novelty detection.
Our workshop will only be accepting full papers.
Full paper submissions to the workshop must be of 8 page papers, with unlimited space for references and supplementary materials, following the CVPR 2026 style and formatting guidelines. The review process is double-blind and there is no rebuttal. Submissions must not have been previously published in a substantially similar form. Accepted papers will be invited for either spotlight talks or poster presentations. Accepted papers will be published in conjunction with CVPR 2026 proceedings.
Submission deadline: Feb 26, 2026 (Anywhere on Earth) DEADLINE EXTENDED: Mar 02, 2026 (Anywhere on Earth)
Author notification: Mar 20, 2026 (Anywhere on Earth)
Camera-ready deadline: Apr 08, 2026 (Anywhere on Earth)
Workshop: June 3rd or 4th, 2026
Please check the Call for Papers page for more details.
Challenges
We retain two complementary tracks that map to pressing industrial needs:
Category 1 — Adapt & Detect: Robust & Efficient Anomaly Segmentation
Category 2 — VLM Anomaly Challenge: Few-Shot Logical & Structural Detection.
Both tracks are designed to (i) stress-test under realistic conditions (domain shift, limited supervision, resource limits) & (ii) incentivize accuracy, robustness, and efficiency.
Please check the Challenge page for more details.
Organizing Team
Philipp Seeböck
MedUni Vienna
Latha Pemula
AWS AI Labs
Singapore Manage-ment University
Toby Breckon
Durham University
Samet Akcay
Intel
Program Committee
Alex Mackin — Amazon, USA
Arian Mousakhan — University of Freiburg, Germany
Ashwin Vaidya — Intel, USA
Botond Fazekas — Medical University of Vienna, Austria
Branko Mitic — Medical University of Vienna, Austria
Brian Isaac-Medina — Durham University, United Kingdom
Giacomo Boracchi — Politecnico di Milano, Italy
Giuseppe Morgese — Medical University of Vienna, Austria
Hana Jebril — Medical University of Vienna, Austria
Jan-Hendrik Neudeck — MVTec, Germany
Jiawen Zhu — Singapore Management University, Singapore
Marzieh Oghbaie — Medical University of Vienna, Austria
Meltem Esengönül — Medical University of Vienna, Austria
Mohammed Kamran — Medical University of Vienna, Austria
Neelanjan Bhowmik — Durham University, United Kingdom
Peng Wu — Northwestern Polytechnical University, China
Ronald Fesco — Medical University of Vienna, Austria
Sassan Mokhtar — University of Freiburg, Germany
Silvio Galesso — University of Freiburg, Germany
Taha Emre — Medical University of Vienna, Austria
Varun Kotte — Adobe
Wenjun Miao — Beihang University, China
Yona Falinie Abd. Gaus — Durham University, United Kingdom
Yunkang Cao — Huazhong University of Science and Technology, China
Zhiwei Yang — Xidian University, China
Contact: vand-cvpr2026@googlegroups.com