Physiological phenotyping tasks are usually complex, and combining data coming from various sensors need to be set into a meaningful context. Moreover, certain devices such as hyperspectral imaging sensors are producing a massive amount of raw data, which then again poses an additional challenge for handling and analyzing.
Nowadays, the application of artificial intelligence increasingly is used to analyze large multidimensional data cubes and to link recorded data to biological information. While machine learning algorithms can be trained to recognize complex physiological traits, e.g. the response to plant diseases. still human intelligence is required to ensure that the artificial intelligence algorithm really learns the targeted traits. An example from our DePhenSe project demonstrates the capabilities but also reveals the challenges.
Schramowski P., Stammer W., Teso S., Brugger A., Herbert F., Shao X., Luigs H.-G., Mahlein A.-K., Kersting K. (2020) Making deep neural networks right for the right scientific reasons by interacting with their explanations. Nat Mach Intell, 2, 476–486.