COCONUT MATURITY LEVEL DETECTION SYSTEM

COCONUT MATURITY LEVEL DETECTION SYSTEM

Sinchan Kunder 1, Prof. Sowmya M S 2

1 Student, Department of MCA, Bangalore Institute of Technology, Karnataka, India

2 Professor, Department of MCA, Bangalore Institute of Technology, Karnataka, India

---------------------------------------------------------------------***--------------------------------------------------------------------

Abstract:

Coconuts are very nutritious but very difficult to carve. We, therefore, present an approach to defining a deep source of SLOV2 learning that can segment coconuts automatically. Overview SOLOv2 is the state-of-the-art one-stage detector that performs the task of precise real-time instance segmentation. Materials and Methods We had 344 images for training and 98 for validation, while the test was on 53 novel classes using Python 3 on Google Colab using a GPU, T4. The tests showed that SOLOv2 delivers very good performance for the recognition of ripeness in coconut. This is evidenced by the Average Precision (AP) of 63.80%, the Average Recall of 82.10%, and the overall mAP of 63.80%, yet without symptoms of overfitting. This further enhances matters of efficiency and costs in processing coconuts and benefits farmers and processors by reducing expenses incurred on labor.

Keywords - Coconut, Effieciency, Instance segmentation, SOLOv2, Performance metrics (Average Precision, Mean Recall, mAP)