October 2020 ELFT Updates

[Evaluation of Latent Friction Ridge Technology (ELFT)]

We've been silent for a bit, but wanted to provide some updates:

• We still feel on track to launch ELFT by the end of fall.
• Timing: By popular request, we've extended timing requirements for template creation for palm and other non-distal regions. We will not be extending timing any further. 
  • We anticipate allowing submissions of "fast" and "slow" algorithms at some point in the future, but not at launch, so pick your best implementation for now.
• Merge templates: We've added a new required method, ExtractionInterface::mergeTemplates(). This method takes one or more templates returned from ExtractionInterface::createTemplate() and requires returning a single template. The use case for this method is when NIST has more than one transaction for a particular subject and does not provide both transactions simultaneously to ExtractionInterface::createTemplate().
• Quality map: Discussions with several latent practitioners indicated the importance of a ridge quality/confidence map. We've added a member to EFS (rqm) for you to export a ridge quality map.
  • We may also provide quality map information as input to ExtractionInterface::createTemplate(), so test this value if that information is important to your algorithm.
• The ELFT agreement form has been posted. This will need to be completed, scanned, emailed, and physically mailed with proof to be valid while NIST continues to operate remotely. A picture of the mailing envelope or a tracking number suffices for proof.
• If you have a submission ready now, even if it is not finalized, we'd love to have it as we finish our testing infrastructure. If you have a finalized submission, we can also begin to build your enrollment database.
  
  • For test submissions, we suggest using a different version numbering scheme (0xFFF0, for instance).
  • Submit an agreement and follow the current validation instructions for submission if you’re interested.
  • Note that you'll still need to rebuild and resubmit once the version number in the API changes to 1.0, so please don't consider your submission final unless you are absolutely certain there will be no changes to your feature extractor.
• Finally, we cannot stress enough how thankful we would be for you to implement the optional methods ExtractionInterface::extractTemplateData() and SearchInterface::extractCorrespondence(). The information provided will be invaluable to future forensics research at NIST.

We're looking forward to seeing your submissions soon!

Thanks,
-Greg (@nist.gov)