Researchers trained a in Deep learning brings speed, accuracy to the life sciences.

Researchers trained a neural network with random plant samples, shown here, from a large French herbaria data set. Next, the network was trained on a much smaller Costa Rican database. The French and Costa Rican species are different, but the model identified Costa Rican species with 80 percent accuracy. This training process, called transfer learning, can help taxonomists identify plants in countries that have small numbers of plant images in digital collections. Photograph: Reprinted from Carranza-Rojas J et al. 2017, Going deeper in the automated identification of herbarium specimens. BMC Evolutionary Biology.