Photographs And Animal Identification: The Promise Of Sloop

Engaging in effective conservation requires precise information about the life history and population ecology of animals, and conserving rare and endangered species can be tricky because of their very rarity. Intrusive methods such as “Capture-Mark-Recapture” are increasingly becoming outdated, as conservationists hope to move towards photo identification and camera traps that are potentially cheaper to gather data from, and far less intrusive. While gathering photos from camera traps may be cheaper than trapping animals themselves, analyzing the photo data is the time intensive part. Searching databases of animal photos manually can be incredibly resource intensive. A database of just 10,000 photos, at just 10 seconds per comparison, would take about 15 person-years to analyze. According to researchers, “the process is difficult to automate and extend and may lead to imprecise quantitative analysis.”

We live in a time of automated photo recognition, however, and Sloop has emerged as a potential solution, using visual features and textual metadata, as well as human input to deliver “high performance” results. The purpose of this particular study / review was to outline the Sloop system architecture, workflows, and algorithms using a handful of endangered species as a case study. They tried a variety of methods, including refining algorithms based on new information, as well as crowdsourcing searchers using Mechanical Turk, which proved to be very successful, with the best matchers producing a 99.96% recognition rate and answering over 1,000 tests in a day. By combining such natural human recognition power with the machine pattern recognition of Sloop allows it to get more and more powerful the more it is used. The researchers describe one particular case study with skinks, where Sloop was able to identify double identities in an image database, as well as identify individual skink’s movements between study sites, adding to knowledge of movement behavior. The researchers found that sloop was able to match 96% of Otago skinks, while it had a 99% match rate with Grand skinks.

Truly, Sloop shows a great deal of promise as a conservation data collection and comparison tool, especially when paired with human abilities. The researchers note that it’s available as a web app and a standalone app, and those curious can check it out here. For curious conservationists, even armchair ones, it’s worth reading up on.