Slides were presented at the IAI 97th International Conference, Phoenix, Arizona. The material in this presentation covers much of the same ground the Eval #2 report.
Slides that were presented at the IAI 97th International Conference, Phoenix, AZ, on latent fingerprint matcher fusion. Results show that when fusing two matchers, representative performance deltas are in the range of 6-8 % points when the input is derived from a single latent image; and up to 15% points when the input comes from two distinct images from the same subject. These results are illuminating for estimating potential future potential performance gains.
May 2012: ELFT-EFS Evaluation #2 Final Report Released
This report describes the design, process, results, and conclusions of ELFT-EFS Evaluation #2; an accuracy test of latent fingerprint searches using features marked by experienced human latent fingerprint examiners, in addition to automatic feature extraction and matching (AFEM).
March 2011: ELFT-EFS Evaluation #1 Final Report Released
This report describes the design, process, results, and conclusions of ELFT-EFS Evaluation #1; an accuracy test of latent fingerprint searches using features marked by experienced human latent fingerprint examiners, in addition to automatic feature extraction and matching (AFEM).
April 2010: ELFT-EFS Evaluation 2 Announced
ELFT-EFS Evaluation #2 started on 21 June 2010. The evaluation was open to all developers of latent fingerprint identification systems. All results were made public and anonymous participation was not be permitted. The latent and exemplar images and features in Evaluation #2 were similar but not identical to those in Evaluation #1. The API and test protocols used by for Evaluation #2 closely followed (but are not limited to) those specified in Evaluation #1. A major change from Evaluation 1 was that participants could specify the subtests on which NIST executed their SDKs. The purpose of ELFT-EFS Evaluation #2 was the continued performance evaluation of AFEM (image-only) and latent fingerprint examiner assisted (image+manual feature markup) based latent fingerprint identification technologies. Technological progress was measured with respect to prior evaluations with the expectation that participants in previous tests may incorporate lessons learned from previous results and/or miss analysis conducted at NIST, while opening registration to new participants.
January 2010: ELFT-EFS Evaluation 1 Preliminary Report Available
Most, but not all of the data processing for ELFT-EFS Evaluation #1 is completed and is reported in this document . Some of the data processing, most notably the multi-encounter galleries, was completed and included in the final report. This document provides an opportunity to present the results available to date without awaiting additional processing and the final report.
July 2009: ELFT-EFS Evaluation 1 Announced
The ELFT-EFS Evaluation 1 started on 2 August 2009 and is limited to all ELFT-EFS Public Challenge participants that submitted results by the 28 June 2009 ELFT-EFS Public Challenge deadline. Anonymous participation in ELFT-EFS Evaluation 1 were not be permitted. ELFT_EFS Evaluation 1 evaluated the accuracy of latent matching using features marked by experienced human latent fingerprint examiners. The purpose of this test was to evaluate the current state of the art in latent feature-based matching, by comparing the accuracy of searches using images alone with searches using different feature sets. The features sets included the current IAFIS latent feature set, and different subsets of the Extended Feature Set (EFS) features proposed by CDEFFS . A key result of the test was to determine when human feature markup is effective. Because human markup is expensive in terms of time, effort, and expertise, there is a need to know when image-only searching is adequate, and when the additional effort of marking minutiae and extended features is appropriate. Here is the test plan for the evaluation.
April 2009: The ELFT-EFS Public Challenge Is Now Open
The ELFT-EFS Public Challenge is a practice evaluation: an open-book test on public data to validate formats and protocols. The results are not for substantive analysis, and participants will remain anonymous in their results. The public challenge was conducted from 15 April 2009 through 15 June 2009. The ELFT-EFS Public Challenge instructions and datasets were finalized on 15 April.
February 2009: Announcing NIST Latent Fingerprint Workshop 2009
We are pleased to announce NIST Latent Fingerprint Workshop 2009, to be held on March 19 and 20 at NIST. This workshop will present and discuss the results of the recently completed NIST Evaluation of Latent Fingerprint Technologies (ELFT) Phase II -- which evaluated the performance of "lights-out" encoding and matching of latent images. This workshop will also provide a venue for the upcoming NIST Evaluation of Latent Fingerprint Technologies: Extended Feature Sets (ELFT-EFS).
November 2008: NIST Evaluation of Latent Fingerprint Technologies: Extended Feature Sets
We are pleased to announce ELFT-EFS - the FBI-sponsored NIST Evaluation of Latent Fingerprint Technologies: Extended Feature Sets. The purpose of this evaluation is to determine the effectiveness of human latent examiner marked fingerprint features on latent fingerprint search accuracy, specifically with respect to the comparative accuracy of image-only searches, image+minutiae searches, and image+extended feature searches. The attachment outlines the evaluation, which begins with a public challenge problem and workshop, March 19-20, 2009.
As additional information becomes available, updates will be made to this website with notifications sent to interested parties.