Growth Stage Classification on Lettuce Cultivars Using Deep Learning Models

This study explores the application of deep learning models for classifying the growth stages of lettuce cultivars. The main focus of this study is on comparing the performance of the YOLOv10 and DETR algorithms in identifying four distinct growth stages: early, head formation, leafy, and harvesting. By taking advantage of computer vision techniques and a dataset of 173 images, the study aims to compare transformer-based and anchor-based algorithms in determining which one is better. Results indicate that YOLOv10 outperforms DETR, achieving higher precision (87.9% vs. 64.3%), recall (76.9% vs. 48.3%), and F1 score (0.818 vs. 0.552). The findings suggest that anchor-based approach of YOLOv10 is more efficient on growth stage classification of lettuce, likely due to its ability to handle small object detection and its optimized computational efficiency. By enhancing the accuracy of growth stage classification, the study contributes to sustainable agricultural practices, aligning with Sustainable Development Goal 12, which promotes sustainable consumption and production patterns.