A Segment Anything Model for Melon Pruning Based on Diameter

Melon is a horticultural commodity with high economic value and great potential for cultivation. One important technique to enhance fruit quality is pruning, which limits the number of fruits per plant so that photosynthetic energy is focused on selected fruits. This study aims to implement the Segment Anything Model (SAM) to support pruning decisions by segmenting melon images and measuring their diameters automatically. SAM is a pre-trained model designed to generalize across a wide range of image segmentation tasks. To adapt it for melon imagery, the model was fine-tuned using a specific dataset of melon images under three different optimizer settings: Adam, AdamW, and Stochastic Gradient Descent (SGD). The performance of each configuration was evaluated using two standard metrics, Intersection over Union (IoU) and Dice coefficient. The results showed that the best configuration achieved a segmentation accuracy of up to 0.9 on both metrics. These findings indicate that SAM is capable of precisely identifying and measuring melon diameters, thus providing a reliable and efficient decision-support tool for optimizing pruning strategies and improving overall fruit quality in melon cultivation.