Insects are important to both the functioning of the earth’s ecosystem and its impact on human society. The exact identification of a particular type of bag is often not easy and often takes some time. Therefore, the VDC group, together with Dominik Vondrek from the Zoological Institute of the Charles University College of Science, focused on the possibility of automatically creating a certain system using modern technologies.
Today more than a million species of insects are known and within a species one can often observe the known differences based on sex, developmental stages and color. The exact identification is very important. Insects, known for their biodiversity, play an important and often indispensable role in the functioning of individual ecosystems on earth. On the other hand, there are many types of insects that have the origin of various diseases or agricultural crops.
For a certain type of knowledge, it is important to have long-term expert knowledge, and at the same time, it often takes a long time to identify. Therefore, efforts in the past have been made to automate and simplify the detection of insects, which has often been inaccurate and required a lot of input data. With the development of new methods called spinning nerve centers to provide accuracy and solve most problems in automatic insect detection.
The CNN is a practical image recognition tool that has to reduce the input parameters. The chain is automatically converted into an automatic processing in vector form that goes into the neural networks with which it works and is categorized.
As this is a promising method, the scientists focused on first determining the accuracy of automatic insect detection using an entangled neural state. They used three different photographic data sets for their research. The first two sets of data, consisting of images of different species, were used to determine whether the automated system could learn to recognize different groups of beetles and biceps beetles. The remaining two data sets were exchanged to identify individuals at the species level.
The first data set contained 884 individuals from two canals of 11 elk imaged from a frontal view of the head, and the mean here was 92.7%. In beetles (2,936 different individuals of 14 icicles, which were photographed from the dorsal side) the accuracy was 96.1%. In the other two datasets, the researchers tried to test the accuracy of automatically recognized species, which is very similar and difficult to identify even for human experts. In this case 97.3% were spontaneously correctly detected for the three related goldfinch species of the genus Oxythyrea (339 individuals, hilly landscape) and 98.6% for nine larval stalk species (3,845 from a single sample photographed from different perspectives).
The research results are very satisfactory or they have shown that, with the help of modern technologies, it is possible to identify certain types of insects with an accuracy of more than 90%. This was already possible in very small data sets, and the results could be compared with assessments by experienced entomologists. By using the convolution of neural networks it will be possible to fully automate the specific taxonomic system.
The use of a bundle of neural networks in the field of biological diversity is still in its infancy, but can now be observed in a familiar way. The authors point out that with the help of this method in the future, in addition to the identification of the existing species, species not described or so-called hidden species can be discovered, which differ only in a few clear characteristics and are normally recognized, especially DNA.
Valan, M., Makonyi, K., Maki, A., Vondrek, D., Ronquist, F. (2019): Automated taxonomic identification of insects with expert-level precision using efficient feature transfer from convolutional networks. Systemic Biology 68 (6): 876895.
