Plant image recognition with deep learning: A review

•Fusing features can improve the performance of neural networks.•Data augmentation and transfer learning can deal with insufficient data.•Latest research focus on solving the problem of insufficient data, model light-weighting. Significant advances in the field of digital image processing have been achieved in recent years using deep learning, which has significantly exceeded previous methods. Deep learning allows computers to automatically learn pattern features. Manual extraction of plant image features requires careful engineering and considerable domain expertise, so how to use deep learning technology for plant image identification studies has become a research hotspot. The following three elements are presented in this work: the various neural network structures in plant image recognition and recent research on neural network improvement methods; the way of plant image data collection and processing; three important future development directions. This review summarizes the methods used in the field of plant image recognition in the past five years, providing the latest and most practical ideas for solving problems for researchers in this field.