Dataset for: A web-based app for sunn pest species recognition (Hemiptera: Scutelleridae: Eurygaster) using machine learning

Keywords: convolutional neural networks, pests, taxonomy, automated species recognition

Desription: This dataset was used in a project focused on training a convolutional neural network (CNN) for expert-level identification of Eurygaster species. It comprises 5,072 dorsal habitus images of specimens that were reliably identified by expert taxonomist V. Neimorovets from the collection of the Zoological Institute, St. Petersburg. We prepared 48 image sets—separately for males and females of each Eurygaster species (see Table 1)—using a variety of imaging equipment, including a DSLR camera (Canon 50D), a compact camera (Olympus TG-5 with and without flash), and a smartphone (Redmi Note 9 Pro):

Eurygaster austriaca (Schrank, 1776)

Eurygaster dilaticollis Dohrn, 1860

Eurygaster integriceps Puton, 1881

Eurygaster laeviuscula Jakovlev, 1886

Eurygaster maura (Linnaeus, 1758)

Eurygaster testudinaria (Geoffroy, 1785)

Likewise, we generated 12 sets of dorsal habitus images using Canon 50D only for each sex of the following true bug species, which resemble Eurygaster spp., to serve as an “outgroup” for CNN training. These species belong to the families Scutelleridae and Pentatomidae:

Aelia acuminata (Linnaeus, 1758)

Dolycoris baccarum (Linnaeus, 1758)

Peribalus strictus (Fabricius, 1803) (= Holcostethus vernalis (Wolff, 1804))

Odontotarsus purpureolineatus (Rossi, 1790)

Piezodorus lituratus (Fabricius, 1794)

Staria lunata (Hahn, 1835)

All images were organized into a total of 60 archive files to facilitate downloading (see Table 1). The abbreviations used in the archive names are as follows:

***_Canon50D – images were taken with a DSLR camera Canon EOS 50D equipped with Sigma 105mm f/2.8 macro lens and Yongnuo YN24EX Macro Flash;

***_Olympus_Flash – images were taken with an Olympus Tough TG-5 with its built-in flash;

***_Olympus_no Flash – images were taken with an Olympus Tough TG-5 without its built-in flash;

***_ Olympus_Redmi – images were taken with a smartphone Redmi Note 9 Pro;

***_M_*** – sets of males;

***_F_*** – sets of females.

Table 1. Number of photographed specimens used in this study:

Species

Canon 50D

Olympus Tough TG-5 with flash

Olympus Tough TG-5 without flash

RedMi Note 9 Pro

Eurygaster austriaca, male

100

100

100

100

Eurygaster austriaca, male

100

100

100

100

Eurygaster dilaticollis, male

75

75

75

75

Eurygaster dilaticollis, female

82

82

82

82

Eurygaster integriceps, male

100

100

100

100

Eurygaster integriceps, female

100

100

100

100

Eurygaster laeviuscula, male

7

7

7

7

Eurygaster laeviuscula, female

5

5

5

5

Eurygaster maura, male

100

100

100

100

Eurygaster maura, female

100

100

100

100

Eurygaster testudinaria, male

100

100

100

100

Eurygaster testudinaria, female

100

100

100

100

Outgroup

Aelia acuminata, male

100

–

–

–

Aelia acuminata, female

100

–

–

–

Dolycoris baccarum, male

100

–

–

–

Dolycoris baccarum, female

100

–

–

–

Peribalus strictus, male

100

–

–

–

Peribalus strictus, female

100

–

–

–

Odontotarsus purpureolineatus, male

100

–

–

–

Odontotarsus purpureolineatus, female

100

–

–

–

Piezodorus lituratus, male

100

–

–

–

Piezodorus lituratus, female

96

–

–

–

Staria lunata, male

100

–

–

–

Staria lunata, female

100

–

–

–