The idea of developing a networked mobile application to recognize species of macrofungi by image processing is not new, but the possibilities offered by state-of-the-art Deep Learning Convolutional Neural Networks (CNNs) for image recognition have put the feasibility of this concept under a new light.
MIND.Funga app has the intention to offer interactivity: the user takes a photograph of macrofungi using the built-in camera of their mobile Phone, and MIND.Funga then classifies using a CNN trained on a central library of thousands images stored in a remote database. MIND.Funga app automatically determines the contours of the macrofungi and uses visual recognition software to find a match for it in the database. Results are returned to the user in a few seconds, depending on the speed of the network connection. Next, MIND.Funga brings up high-resolution images of the fungi, along with images of the species list with a percentage of identification possibilities. The app also supplies background information on the species and its geographic distribution. When the identification is not straightforward, MIND.Funga app users dig into other related images in its database, such as mushroom shape or details patterns. In the end, it is up to the user to make the determination of the species, which reinforces mycological learning. Once a user successfully identifies a fungi, his or her photograph and accompanying GPS location data are automatically uploaded into MIND.Funga app’s database, contributing to the work of a community of scientists who are using the stream of data to map and monitor how the abundances and geographic ranges of different macrofungi species are changing through time and as a function of climatic change.
We are starting a collaboration network in order to develop a mobile phone application aimed at offering users an identification system of macrofungi species from South America. One of the first steps in order to develop this ID app is to build a big database of photos of different groups of macrofungi. In collaboration with the LAPIX – UFSC team (http://www.lapix.ufsc.br/), this image database will be used to train CNNs to recognize species and offer a list of identification candidates of macrofungi ranked by probability for the app’s users.
If you are a taxonomist/parataxonomist, we would like to invite you as a formal collaborator with your fantastic pics (with Copyright of course). Do you think that it is possible?
As mentioned before, we are starting. So, at this time we are forming a collaboration network, then we are going to start with database and training the software. Our idea is to offer diversity information for people in general (curious, students, teachers, nature lovers, etc). Firstly, we are offering possibilities of IDs and, consequently, we have the opportunity to research in a broader way, counting with the collaboration of the users.
Despite the app being in a test phase, the MIND.Funga team recently released a protocol to capture macrofungi images that will help users with the app. This protocol is available for free in an E-book format and gives essential tips that will make your photos scientifically meaningful and usable to train neural networks for the app. Click here to download the bilingual E-book.
We would be very happy to have you in our team. It is easy, you only need to send your fungi pictures with basic information. Please, send an e-mail to email@example.com and we can find the better way. Here we will work on treating the pictures for the database.
We hope to hear from you as soon as possible!