GES: INTELLIGENT SYSTEM FOR DETECTION AND CLASSIFICATION OF DEFECTS IN GRANITE PLATES
Palavras-chave:
ornamental rocks, granite, yolov5, deep learningResumo
According to information from BANDES (Development Bank of Espírito Santo) in 2021,
Espírito Santo was responsible for 82% of brazil's marble and granite exports. This is one of the most
relevant industries of the Economy of Espírito Santo, representing about 7% of its GDP (Gross Domestic
Product). As a comparison, in November 2021 alone, Brazil exported 221,800 tons of ornamental rocks for
$138.1 million, accumulating 2.21 million tons of exported rocks in the year, totaling $1.2 billion in the
period. Using the process of processing granite sheets, there are many manual activities, such as inspection,
cataloging, photography, and registration. Such volume of interactions makes the process slow, and prone
to errors that can influence the last price of the product. The applicability of artificial intelligence techniques
in these processes is observable, and it is possible to present results from the use of high-resolution digital
images, initially fitting the classification and organization of images in a dataset applicable to machine
learning techniques. The present work aims to build a classification system of these granite plates, called
"GES System", trained from the dataset with images provided by the companies participating in the project
and that will be able to identify the rock class, and its defects. The classification process will be done using
YoloV5 artificial intelligence tools. The uses of other languages for the complete completion of the project
and release for the use of the system in a production environment are planned for other stages. Among the
benefits of process automation are cost reduction, agility in the process of identifying defects and
standardization of classification.