Tattoo Recognition Technology-Challenge (Tatt-C)

 

Tatt-C Dataset

November 2, 2016:
NIST has announced a follow-on activity to Tatt-C, which is a large-scale, sequestered evaluation called Tatt-E.

Tatt-C Publications

September 15, 2015: 
The first public report on Tatt-C has been published as NISTIR 8078 – Tattoo Recognition Technology – Challenge (Tatt-C) Outcomes and Recommendations.

March 24, 2015:
Mei Ngan and Patrick Grother.  Tattoo Recognition Technology - Challenge (Tatt-C): An Open Tattoo Database for Developing Tattoo Recognition Research.  In International Conference on Identity, Security and Behavior Analysis (ISBA), pp.1-6, 2015.

Tatt-C Ongoing Updates

June 17, 2015: Tatt-C workshop proceedings posted
The online proceedings of the Tatt-C workshop held on June 8, 2015 are now available here.

Scope

The Tattoo Recognition Technology – Challenge (Tatt-C) is being conducted to challenge the commercial and academic community in advancing research and development into automated image-based tattoo matching technology. The activity will assess the capability of image-based tattoo recognition algorithms to perform detection and retrieval of tattoos, with the goals to determine which algorithms are most effective and whether any are viable for the following operational use-cases: 1. Tattoo Similarity – matching visually similar or related tattoos from different subjects; 2. Tattoo Identification - matching different instances of the same tattoo image from the same subject over time; 3. Region of Interest - matching a small region of interest that is contained in a larger image; 4. Mixed Media - matching visually similar or related tattoos using different types of images (e.g. sketches, scanned print, computer graphics, or natural images); 5. Tattoo Detection - detecting whether an image contains a tattoo or not.

Interested Parties

Please contact NIST if:

a) You are a developer of tattoo matching algorithms or have an interest in developing such a capability.

b) You represent an organization possessing suitable tattoo datasets that might be valuable to our effort.

c) You have an operational interest or need for image-based matching of tattoo images.

Background

Tattoos have been used for many years to assist law enforcement in the identification of criminals and victims and for investigative research purposes.* Historically, law enforcement agencies have followed the ANSI-NIST-ITL 1-2011 standard to collect and assign keyword labels to tattoos. This keyword labeling approach comes with drawbacks, which include the limitation of ANSI-NIST standard class labels to describe the increasing variety of new tattoo designs, the need for multiple keywords to sufficiently describe some tattoos, and subjectivity in human annotation as the same tattoo can be labeled differently between examiners. As such, the shortcomings of keyword-based tattoo image retrieval have driven the need for automated image-based tattoo recognition capabilities.

Structure of Tatt-C

Tatt-C is structured around problems that are designed to challenge the commercial and academic community in advancing research and development into automated image-based tattoo recognition technology. While some research and commercial capability is available, tattoo recognition is not a mature industry. There is no common test data and use cases to evaluate and develop systems for next generation government applications. To address this shortcoming, the Tatt-C dataset was developed as an initial tattoo test corpus that addresses use cases derived from operational scenarios provided by the FBI's Biometric Center of Excellence (BCOE).

The Tatt-C dataset consists of still images of tattoos captured operationally by law enforcement agencies. The operational nature of this corpus imposes challenges on traditional image retrieval methodologies given the large variation in capture environment/process and tattoo content/quality. The following are examples of such challenges represented in the Tatt-C dataset:

• Inconsistent image lighting and scale
• Occlusions due to clothing and different image backgrounds
• Different tattoo background or embellishments around primary tattoo content
• Blended images or multiple tattoos in a single image
• Inconsistent orientation of body/appendages and images
• Extremely faded tattoos
• Ambiguous or unfamiliar abstractions (difficult or impossible for different people to view and interpret consistently)

The Tatt-C dataset provides a basis for objectively measuring and comparing tattoo recognition capabilities, with partitions focused on but not limited to the following use cases:

• Tattoo Similarity. What is the retrieval performance for finding visually similar or related tattoos from different subjects?
• Tattoo Identification. What is the retrieval performance for finding different instances of the same tattoo image from the same subject over time?
• Region of Interest. What is the retrieval performance for finding a small region of interest that is contained in a larger image?
• Mixed Media. What is the retrieval performance for finding visually similar or related tattoos using different types of images (e.g. sketches, scanned print, computer graphics, or natural images)?
• Tattoo Detection. What is the performance for detecting whether an image contains a tattoo or not?

Reference

Please cite references to the Tatt-C dataset as:

Mei Ngan and Patrick Grother.  Tattoo Recognition Technology - Challenge (Tatt-C): An Open Tattoo Database for Developing Tattoo Recognition Research.  In International Conference on Identity, Security and Behavior Analysis (ISBA), pp.1-6, 2015.

Important Dates:

September 23, 2014 – February 6, 2015: Phase 1 participation window 

February 7 – May 4, 2015: Phase 2 participation window 

June 1, 2015: Registration deadline to attend Tatt-C workshop 

June 8, 2015: Tatt-C workshop at NIST, Gaithersburg, MD 

Subscribe to the Tatt-C mailing list to receive emails when announcements or updates are made – tattooinfo-request [at] nist.gov (subject: subscribe) (subscribe).

---

*A sentence was removed on May 19, 2016 based on feedback we received which indicated that the previous text did not accurately convey the intent of the project.

Tatt-C links

Past and Current publications
Tatt-C Workshop Proceedings (June 8, 2015)