Analysis on localization and prediction of depth chili fruits images using YOLOv5

Chili fruits are essential ingredients that Malaysians consider essential for cooking. Adding chili to a dish used to produce a second fiery flavour. In south-western Ecuador, one of the most important plant-growing regions on the American continent at the time. As a result, this provides evidence that people were using chili as an additional food element as early as 600 years ago. A traditional method of picking chili is common, but it is less precise and time-consuming. Incorrect picking and grading will cause the harvesting process to take longer. The advancement of computer vision and pattern recognition has demonstrated its effectiveness in image recognition. Because of their simplicity and low complexity, 2 dimensional images are frequently used in image recognition. As a result, advancement in automated picking systems with object detection is common. However, due to a lack of image information, such as depth, 2D images are thought to be difficult to identify the growing stages or maturity level of chili fruits. Object detection is prevalent for determining the localization and category of objects. One of the well-established methods such you look only once (YOLO) has widely used is in object detection. To anticipate this effort, the fast, reliable and able to recognize small object, YOLOv5 is proposed to localise and predict the category of chili fruits which allows the process to determine a chili's form and categories based on its colour. The proposed model is able to differentiate and localize the position as well as the colour of chili fruits with above 94% in average. Hence, our achievement has proven its effectiveness and becomes our greater goal of developing an autonomous chili fruit picking robot which could help farmers or agricultural sectors to reduce their labours during the grading process.