Kaputt: A Large-Scale Dataset for Visual Defect Detection

  • Sebastian Höfer1
  • Dorian Henning1
  • Artemij Amiranashvili1
  • Douglas Morrison1
  • Mariliza Tzes1
  • Ingmar Posner1,2
  • Marc Matvienko1
  • Alessandro Rennola1
  • Anton Milan 1


    • 1                   2 University of Oxford

Paper        Dataset
Supplementary Material

Abstract

We present a novel large-scale dataset for defect detection in a logistics setting. Recent work on industrial anomaly detection has primarily focused on manufacturing scenarios with highly controlled poses and a limited number of object categories. Existing benchmarks like MVTec-AD (Bergmann et al., 2021) and VisA (Zou et al., 2022) have reached saturation, with state-of-the-art methods achieving up to 99.9% AUROC scores. In contrast to manufacturing, anomaly detection in retail logistics faces new challenges, particularly in the diversity and variability of object pose and appearance. Leading anomaly detection methods fall short when applied to this new setting. To bridge this gap, we introduce a new benchmark that overcomes the current limitations of existing datasets. With over 230,000 images (and more than 29,000 defective instances), it is 40 times larger than MVTec and contains more than 48,000 distinct objects. To validate the difficulty of the problem, we conduct an extensive evaluation of multiple state-of-the-art anomaly detection methods, demonstrating that they do not surpass 56.96% AUROC on our dataset. Further qualitative analysis confirms that existing methods struggle to leverage normal samples under heavy pose and appearance variation. With our large-scale dataset, we set a new benchmark and encourage future research towards solving this challenging problem in retail logistics anomaly detection. The dataset is available for download under https://www.kaputt-dataset.com.

Citation

Dataset Download

Please fill out the following form to obtain access to the dataset. Once completed, you will receive detailed information on how to download the dataset.

Terms of Use - Kaputt Defect Dataset (KDD)

These terms govern your use of this Kaputt Defect Dataset ("KDD").

  1. You acknowledge and agree that the KDD is subject to the Creative Commons (CC BY-NC-ND 4.0) license. Where the following terms and conditions (the "Superseding terms") conflict with the Creative Commons (CC BY-NC-ND 4.0) license, the Superseding Terms take precedence.
  2. You agree to use this KDD solely for education and scientific research in the field of computer vision (the "Authorized Purpose").
  3. Amazon is providing the KDD free of charge, on an "as is" basis, and solely for the Authorized Purpose. Amazon makes no representations or warranties of any kind, express or implied, with respect to this KDD, including as to third-party approval, endorsement, or non-infringement.
  4. You expressly agree that your use of this KDD is at your sole risk.
  5. If you are accessing or using this KDD on behalf of or for the benefit of an entity or institution, you represent and warrant that you are authorized to access these Terms on the entity or institution's behalf.
  6. This KDD may include images that incorporate third-party intellectual property.
    1. You are solely responsible for determining if your use of third-party intellectual property contained in this KDD is fair, nominative, or otherwise lawful. Amazon makes no representations in this regard. You agree to seek independent legal advice on this issue as appropriate.
    2. You acknowledge and agree that the presence of third-party intellectual property in this KDD does not imply approval or endorsement by the intellectual property owner(s), or a representation that this intellectual property is available for lawful use.
    3. You agree not to use this KDD in a manner that is likely to cause confusion or otherwise infringes third-party intellectual property.
  7. You agree not to remove these Terms from this KDD.
  8. Any dispute or claim relating to your use of this KDD will be adjudicated in the state or Federal courts in King County, Washington, and you consent to exclusive jurisdiction and venue in these courts.
  9. You agree to indemnify Amazon, its affiliates, agents, servants, officers, directors and employees, from and against all claims, causes of action, liabilities, suits, administrative actions, damages, costs, or expenses (including reasonable attorneys' fees), whether involving a third-party claim or a direct claim between you and Amazon, that arises from or is related in any way to your use of this KDD.
  10. To the full extent permissible by law, Amazon disclaims all warranties, express or implied, including, but not limited to, implied warranties of merchantability and fitness for a particular purpose.
  11. To the full extent permissible by law, Amazon will not be liable for any damages of any kind arising from your use of this KDD, including, but not limited to direct, indirect, incidental, or punitive damages.
  12. By accepting these terms of use, you confirm that you are at least eighteen (18) years of age or older.
By submitting this form, you acknowledge that you have read our Privacy Notice.

✅ Dataset Download Request Submitted Successfully!

You will shortly receive an email with detailed instructions as to how to access the dataset. Please also make sure to check your spam folder.

For any questions or concerns, please contact us at @amazon.com providing your submission ID.

🔒 Security Verification

Please complete the CAPTCHA below to continue