MIND.Funga App: images dataset from Neotropical macrofungi used to train an artificial neural network to recognize fungal species

Description

Macrofungi produce reproductive structures (spore-bearing), the sporomata, visible to the naked eye. These sporomata have variable shapes, sizes, and colors. They are commonly known as mushrooms, puffballs, earthstars, jelly fungi, and bracket fungi. Informative photos can often be enough to recognize species or taxonomic groups of these macrofungi. Therefore, we built a database with standard and good-quality images for species recognition through the MIND.Funga application (MIND.Funga App). The MIND.Funga App does macrofungi image recognition based on deep learning neural network (DLNN) and suggests species names with confidence ratings. This second version of the database used by MIND.Funga App contains 17,467 images that were manually curated and resized (all images are squared). All images are separated by taxons (more than 500 species) in individual directories named with the current accepted scientific name. When species identification was not possible, the directory was named with the specimen's genus. The images were gathered and compiled from several sources, including collaborators from Facebook groups, citizen scientists, partner researchers, scientific papers, and from MIND.Funga research projects and personal data from MIND.Funga members. For the proper operation of the MIND.Funga app, it is necessary that the database used to train the DLNN be constantly supplied and composed of an expressive number of good-quality images and a representative number of species. Thus, it is expected that new versions of the database will be published in the near future. This research is supported by FAPESC/CNPq (PRONEM) 2020TR733. For more information on the collaboration net and the project, see https://mindfunga.ufsc.br/.

Institutions

Categories

Funding

Licence