Project duration: 2014-2019
The fast and accurate identification of plants is of great importance to many people. Smartphones and tablets in combination with cutting-edge artificial intelligence methods now allow this task to be solved in a matter of seconds.
Fewer and fewer people today are able to reliably identify a large number of species and put their occurrence into an ecological context – herbaceous plants in particular are difficult for laypeople to distinguish. To the untrained eye, plants form a uniform green mass. Even a ” meadow of flowers” is perceived by many people only as “grass”. But even for trained biologists, precise plant identification is often a major hurdle.
This development is precarious for nature conservation and land use. As knowledge about animals, plants and ecological interrelationships dwindles, so does the willingness of the population to get involved in nature conservation and environmental protection issues. How can one campaign for the protection of biodiversity if this knowledge is practically non-existent among the population? The basis for protection, maintenance and development measures is a continuous inventory of flora and fauna. How can this be ensured if the number of experienced species experts is continuously decreasing?
In the population and especially among children and young people, interest in digital media has been increasing for years. Many people can no longer imagine their lives without them. Since the introduction of Apple’s iPhone in 2007, smartphones have become constant companions and an indispensable tool for many users. At the same time, we are witnessing rapid technical development in the field of machine learning and artificial intelligence. These technical developments are already influencing the way we learn and acquire knowledge.
Is it possible to bring classical plant identification into the digital age with this?
Artificial intelligence describes the ability of machines to autonomously perform tasks based on algorithms. This ability can be based on programmed processes or generated by machine learning (ML). In machine learning, an algorithm learns to perform a task autonomously through repetition.
As part of the Flora-Incognita project, computer scientists have trained deep learning artificial neural networks (so-called Convolutional Neural Networks), whose structure and functioning are inspired by vertebrate brains. An artificial neural network consists of different nodes (neurons). They take in information from other neurons or from outside, modify it and then pass on a result to the next neuron. There are several layers of these neurons in a network: the input layer, the output layer, and “hidden layers” in between. Raw data is captured by neurons in the input layer and passed on to the next layer down. State-of-the-art deep learning networks allow a large number of hidden layers to be trained efficiently, thereby dramatically increasing the performance of the network.
Comparison between biological and artificial neurons and neural networks
For an artificial intelligence, a plant consists of characteristic corners, edges, planes and patterns.
For plant identification, millions of plant images with associated species names provide this raw data. The network first “learns” to recognise simple concepts such as corners and edges, and increases this complexity in the further course via surfaces, patterns, until flowers or leaves are composed from these elements near the output layer. By using these deep learning artificial neural networks, Flora Incognita does not need any manual steps to recognise and extract plant features. The network independently learns which concepts are relevant for classifying plants in an extensive and complex training process.
Among other things, this requires a large amount of representative and reliable training data. Training data was collected as part of a large-scale citizen science project using the Flora Capture app. Based on these images, a wide variety of architectures of deep learning neural networks were evaluated for the Flora Incognita app and a hierarchical structure consisting of several such networks was designed.
In order to bring the Flora-Incognita app to an application that can be reliably used by very different people throughout the year, the research team was confronted with various challenges:
Five different types of rannunculus that an artificial intelligence must learn to distinguish.
Addressing these challenges has led to several innovative approaches in both computer science and botany, which are unique in their combination in the Flora Incognita app:
The “Flora Incognita” app developed as part of the project impressively shows how easy plant identification can be today. With the camera of your smartphone you take a picture of the flower, and if necessary also of the leaf, and in a matter of seconds you receive a suggestion for the name of the plant as well as further information. The app thus allows easy and safe plant identification for people of any age and any level of prior knowledge.
The Flora Incognita app was already released during the ongoing research project and has been publicly available on Google Play, the Apple AppStore and in the Huawai AppGallery since spring 2018. By the end of the project in 2019, the app had already been installed 2 million times. Not only laypeople use Flora Incognita, but also botanical experts rate the app extremely positively and recommend it. According to experts at the Federal Agency for Nature Conservation (BfN), the Flora Incognita app has the potential to become the most widely used tool for identifying and mapping German flora. Schools, universities and other educational institutions are already successfully using Flora Incognita for teaching purposes.
In the research project “Flora Incognita ++”, many other questions are now being studied using the data from the Flora Incognita App: When do which species flower? How strongly do the characteristics of a plant species vary? What are the connections to climate change and the type of land use?
More information on the follow-up project can be found on the website:
https://floraincognita.com/wp-content/uploads/2024/10/DSC00396-3.jpg
2048
1536
Anke Bebber
https://floraincognita.de/wp-content/uploads/sites/4/2022/11/flora-incognita_logo_horizontal_RGB-e1669665654345.png
Anke Bebber2024-10-24 10:27:252024-10-24 10:27:25Study on identification accuracy: Flora Incognita reaches 98.8 %
https://floraincognita.com/wp-content/uploads/2024/10/Ranunculus_auricomus_Geschwenda_Wiese_IMG_1173.jpg
1382
1842
Anke Bebber
https://floraincognita.de/wp-content/uploads/sites/4/2022/11/flora-incognita_logo_horizontal_RGB-e1669665654345.png
Anke Bebber2024-10-11 11:15:102024-10-11 11:15:10New publication: Does an artificial intelligence recognize the leaf shape?
https://floraincognita.com/wp-content/uploads/2022/12/AppUse2022_10-1030x773-1.jpg
773
1030
Anke Bebber
https://floraincognita.de/wp-content/uploads/sites/4/2022/11/flora-incognita_logo_horizontal_RGB-e1669665654345.png
Anke Bebber2024-07-29 10:38:112024-07-29 10:38:11How a plant app helps identify the consequences of climate change
https://floraincognita.com/wp-content/uploads/2024/06/Wiese_KK-scaled.jpg
1213
2560
Anke Bebber
https://floraincognita.de/wp-content/uploads/sites/4/2022/11/flora-incognita_logo_horizontal_RGB-e1669665654345.png
Anke Bebber2024-06-26 09:14:232024-06-26 11:00:25New Publication: Species Delimitation 4.0The Flora Incognita Project was a joint project of the Ilmenau University of Technology and the Max Planck Institute for Biogeochemistry Jena. It was funded from 2014 to 2020 by the Federal Ministry of Education and Research, and from 2014 to 2019 by the Federal Agency for Nature Conservation with funds from the Federal Ministry for the Environment, Nature Conservation and Nuclear Safety and the Stiftung Naturschutz Thüringen. The Flora Incognita project was a research project for the implementation of the National Strategy on Biological Diversity. Since 2019, the partners have been working on the follow-up project Flora Incognita++.
