Review: Deep learning on image-based plant phenotyping (Trends Plant Sci)

The development of deep learning brings opportunities to train computers to solve complex questions. Self-driving vehicles are classic examples of an application of deep learning in the real world. However, the large amounts of data that are required for building accurate models and avoiding overfitting problems were previously hard to accomplish in the plant science area. Fortunately, image-based plant phenomics can generate various types of images in a high-throughput way, which accelerates the connection between deep learning and image data. A recent review by Singh et al. summarizes the progress in this field and recapitulates the deployment of deep learning on ICQP (Identification; Classification; Quantification; Prediction). The authors also introduce current popular deep learning frameworks on disease detection and plant architecture identification. The summarized methods will be a valuable guide for computational biologists to refine their research topics and select appropriate deep learning models for corresponding questions. (Summary by Zhikai Liang) Trends Plant Sci. 10.1016/j.tplants.2018.07.004