Detection of tomato plant phenotyping traits using YOLOv5-based single stage detectors

• DOI: 10.1016/j.compag.2023.107757
• Publication Date: 2023-03-09T11:19:31Z
• Authors: Angelo Cardellicchio; Firozeh Solimani; Giovanni Dimauro; Angelo Petrozza; Stephan Summerer; Francesco Cellini; Vito Renò
• Keywords: plant phenotyping; tomato plants phenotyping traits; Tomato plant phenotyping,YOLOv5,Nodes,Fruit,Flower identification; deep learning; 630

Plant phenotyping is the study of complex plant traits to evaluate its status depending on life-cycle conditions. Often, these evaluations are carried out by human operators, and accuracy could be biased their experience skill, especially when dealing with huge amounts data produced high-throughput (HTP) platforms. With rapid development key enabling technologies, HTP only made possible vast available computer vision systems. In this scenario, artificial intelligence algorithms play a role in automation, standardization, quantitative analysis large data. This paper focuses detecting tomato plants using single-stage detectors (either stand-alone or ensemble) based YOLOv5, aiming effectively identify nodes, fruit, flowers challenging dataset acquired during stress experiment conducted multiple genotypes. Results demonstrate that models achieve relatively high scores, considering particular challenges input images terms object size, similarity between objects, color.