Figure 1.

Specimens used to train or test the convolutional neural network. (a) Species with accuracy (images of species in test dataset classified to correct species) on 90% or more. First row: Cylindera germanica, Cychrus caraboides, Calosoma inquisitor, Carabus glabratus, and Lebia chlorocephala. Second row: Broscus cephalotes, Cicindela campestris, Nebria complanata, Carabus problematicus, and Chlaenius nigricornis. (b) Species drawn randomly from the remaining 281 species. First row: Bembidion atrocaeruleum, Amara famelica, Dyschirius politus, Acupalpus meridianus, and Blemus discus. Second row: Bembidion ephippium, Ophonus melletii, Amara lunicollis, Dromius angustus, and Demetrias imperialis