An example of lettuce (Lactuca Sativa) seedling selection using deep learning method for robotic seedling selection system

Abstract

Lettuce is a type of vegetable that is widely cultivated and consumed in our country and in the world. The seedling period, which is the beginning of production, is the most sensitive time for the plant. Starting production with healthy seedlings is an important parameter for quality and efficient production. In this study, a sample program for automatic seedling selection was developed for a robotic system to be used in seedling production. With the developed program, it was aimed to select seedlings with the same degree of maturity in multi-well pots. In this study, Yolo5n was used for the training model. A learning system was established on two types of lettuce (curly salad), and red curly lettuce leaf (lolo-rosso) seedlings. As a result of the training, F1 score was found as 83%; Precision was 100%; Recall was 95%; Precision Recall was 86.7%. The learning rate was 0.0005 for all given images. In view of these data, positive results were obtained for the mentioned method in seedling selection.