Deep Learning on Images and Genetic Sequences in Plants: Classifications and Regressions

Recent progress of deep learning (DL) frameworks has allowed various high-quality classifications and regressions to be developed. Their performances have often exceeded human standards, especially in image diagnoses, so we may be able to reproduce artificial professional eyes for various objectives. Furthermore, recent development of explainable DL technology (or explainable AI (X-AI)), which can visualize the relevance of DL predictions, would allow biological interpretations of the predictions, potentially providing insights that only DL models may be able to recognize. Nevertheless, the application of DL frameworks has still progressed less in plant science than in animal, medical, and social sciences. In this chapter, mainly focused on classification and regression analyses, we introduce the current applications and potential prospects of DL technologies with respect to plant images and genetic sequences. Here, we take as examples taxonomic classification, disease/stress diagnosis, non-invasive prediction, and implementation with automated sorting systems based on plant images, as well as prediction of gene functions, such as expression patterns and protein folding structures, based on DNA or amino acid sequences. For beginners in the use of DL techniques, tips and precautions regarding practical application of DL frameworks are also provided.