I came across an interesting story in the London Telegraph that forms a nice bookend to the proliferation of facial recognition software blog post of last week. The Telegraph article reports that the US Intelligence Advanced Research Projects Activity (IARPA), Incisive Analysis Office, which operates under the Office of the Director of National Intelligence (ODNI) published a solicitation that says it is looking for a system that can quickly identify the location where photos or videos were made.
IARPA, as way of background, "... invests in high-risk/high-payoff research programs that have the potential to provide our [US] nation with an overwhelming intelligence advantage over future adversaries."
IARPA's solicitation, which originally came out in June and then was amended in July [see Amendment 1] reads, in part:
"The Finder Program aims to develop technology that, with the aid of an analyst, geolocates an outdoor image or video taken from anywhere on the land surface of the world, via the use of publicly available information."
The hope is that with a such a system, the location of, say, where a non-geotagged propaganda video was shot or photos captured (for example, on a laptop) at a terrorist/enemy location can be quickly pinpointed.
The solicitation also says that the proposed Finder program:
"... aims to build on existing research systems to develop technology that augments the analyst’s abilities to address the geolocation task. Required technical innovations include the 1) integration of analysts’ abilities and automated geolocation technologies to solve geolocation problems, 2) fusion of diverse, publicly-available, but often imperfect data sources, and 3) expansion of automated geolocation technologies to work efficiently and accurately over all terrain and large search areas. If successful, Finder will deliver rigorously tested solutions for the image/video geolocation task in any outdoor terrestrial location."
"The Finder Program has the following important characteristics:
- Because of the difficulty of the geolocation problem, it is expected that systems in the Finder Program will produce a candidate list of geolocations rather than a single geolocation result.
- It is expected that an analyst is an essential part of the process and will interact with the system, but his/her abilities and time must be well-used.
- There are significant publicly available sources of information that can potentially contribute to the geolocation task. The program seeks to understand the extent to which such information can address the challenge.
- Successful solutions will build on approaches informed by techniques in decision support, optimization, and related fields."
The initial goal seems to be able to analyze/identify a photo in a minute or less, and a video in 3 minutes or less.
The Finder Program, which will proceed in two phases, is expected to begin by January 2012 and end by April 2016; no funding amount has been announced.
The solicitation also states that:
"IARPA is not interested in approaches that only integrate current software. New algorithms and methods must be proposed."
In addition, the solutions offered "are prohibited from using any information about individuals in the geolocation tasks ... [and] ... are prohibited from conducting any analysis on information about individuals in the reference data, e.g., face recognition; person identification or matching; license plate number identification; etc."
However, one can imagine how effective a system—Finder Program 2.0?—could become if such information were to be included.