Computer Vision Assisted Bird–Eye Chilli Classification Framework Using YOLO V5 Object Detection Model

A computer vision-based bird-eye chilli sorting system has become essential due to the rising demand for quality verification and effective sorting of chilli products. The quality of bird-eye chilli, or ‘kantari mulaku’, directly affects the flavour and is a highly sought-after ingredient in many different cuisines around the globe. Computer vision technology-based automated sorting systems can precisely recognize and categorize chillies based on various quality factors like size, shape, colour, and texture. Additionally, computer vision-based sorting systems are perfect for large-scale production facilities because they can constantly run for prolonged periods with little supervision. This paper describes a method for categorizing bird-eye chilli using a 3 DOF robotic manipulator and the You Only Look Once-V5 object recognition algorithm. Images of bird-eye chillies in various orientations and settings make up the dataset used in this research. This dataset was used to train the algorithm, and the model successfully identified and classified bird-eye chilli. The chillies were then grabbed by a robotic manipulator and sorted according to their degree of maturity. The proposed approach obtained an average precision of 0.90 and a mAP of 0.94. Chillies can be graded with high precision, consistency, and efficiency using a robotic manipulator, which boosts output and lowers human error rates. The developed YOLO V5 framework is deployed in Raspberry Pi 4B graphical processing unit, verifying the efficacy. The outcomes of this work show how successfully classifying bird-eye chilli using YOLO V5 can be applied in the food and agricultural industries.