The popular website for nature lovers, iNaturalist.org is launching a deep learning-based app that automatically identifies plants and animals down to the species level.
Founded in 2008 by students at University of California, Berkeley and recently acquired by the California Academy of Sciences, iNaturalist has until now been solely a crowdsourcing site. Users upload a photo, and the community of experts and amateur citizen scientists help identify it – but according to their website, “on average, observations take 18 days to be identified by the community, with half of all observations identified in the first two days.”
To help take the burden off the volunteer experts, the iNaturalist team collaborated with students at the Cornell Lab of Ornithology to develop an app to provide higher quality identifications faster as the community continues to grow.
Using NVIDIA GPUs and cuDNN with the TensorFlow deep learning framework, they trained the neural networks on their massive database of images that have been labeled by the site’s community of experts. Currently, they are able to identify 10,000 different species and are adding new species to the model every 1.7 hours.
An online demo version of the app is now live and you can help train the models by uploading and classifying images.
Besides being a useful tool for naturalist or schoolchildren, the quick identification could also be useful application for law enforcement.
“Let’s say TSA workers open a suitcase and someone’s got geckos,” says iNaturalist’s co-director Scott Loarie. “They need to know whether to arrest someone or not.”