Google He has developed a modelo open sourcewho has called SpeciesNetand that is thought To identify wild animal species analyzing photographs Taken with trap cameras. These are digital cameras connected to infrared sensors who use researchers around the world to study wildlife populations, and collect both the number of species and their state.
These cameras can offer valuable information, since they generate large amounts of data that can take time to several weeks. Therefore, Google launched the Wildlife Insights initiative within the Google Earth Outreach philanthropy program. Wildlife Insights offers a platform, based on Google Cloud, with which researchers can share, identify and analyze images of wild wildlife online, collaborating in the acceleration of the analysis of the data of the trap cameras.
It is in this framework that Speciesnet has developed, which supports many of the initiative analysis tools. The model, according to Google, has trained with around 65 million publicly available images and from organizations such as the Smithsonian Biological Conservation Institute, the Wildlife Conservation Society, the Natural Sciences Museum of North Carolina and the Zoological Society of London.
Speciesnet is capable, thanks to all this, to classify images by assigning one of the more than 2,000 labels that are available, covering from mammals or felines to inanimate objects. From Google they point out that the Speciesnet models will allow both developers and academics and startups that work in areas related to biodiversity can climb their biodiversity monitoring in natural spaces.
The model is Available in Github with an Apache 2.0 licensewhich means that it can be used without restrictions, even commercially. Both for the analysis of fauna images and for the decision -making of biodiversity.
