Advance Deep Learning for Precise Leaf Segmentation and Counting in Tomato and Chili Crops

Abstract

In Pakistan, the agricultural sector has a major economic significance, especially in terms of the production of tomatoes and chilies. Leaf counting is an important aspect of crop analysis since it reveals information about the health of the plants and their yield potential. A limited amount of research has been conducted in Pakistan in the area of segmentation and counting of leaves using advanced deep learning-based methods. The focus of this research is to explore the advanced deep- learning-based methods for the accurate segmentation and counting of tomato and chili crops in Pakistan. This research uses a state-of-the-art instance segmentation model, MaskRCNN, to segment leaf regions and count the number of leaf instances that can be used to determine canopy density, measure leaf area index, and predict potential crop diseases. The results show how our proposed approach can greatly improve segmenta- tion and counting accuracy, opening the door for better crop management and disease control.