MIPDB: A maize image-phenotype database with multi-angle and multi-time characteristics

Abstract

Plant phenomics has become one of the most significant scientific fields in recent years. However, typical phenotyping procedures have low accuracy, low throughput, and are labor-intensive and time-consuming. Large-scale phenotypic collection equipment, on the other hand, is pricy, rigid, and inconvenient. The advancement of phenomics has been hampered by these restrictions. Lightweight picture collection equipment can now be used to capture plant phenotypic data thanks to the development of deep learning-based image identification. For the purpose of training the model, this approach needs high-quality annotated datasets. In this study, we used a handheld camera to gather multi-angle, multi-time series images and an unmanned aerial vehicle (UAV) to create a maize image phenotyping database (MIPDB). Over 30,000 high-resolution photos are available in the MIPDB, with 17,631 of those images having been carefully tagged with point-line method. The MIPDB can be accessed by the general public at http://phenomics.agis.org.cn. We anticipate that the availability of this superior dataset will stimulate a new revolution in crop breeding and advance deep learning-based phenomics research.