Andari Dyah Widowatie, Moechammad Sarosa, Rosa Andrie Asmara
The tobacco business, which includes cigarettes and cigars, is one of the non-oil and gas sectors that provides the most to the country's revenue. Farmers need to figure out strategies to produce the best quality tobacco to ensure profitable income. One of the ways to guarantee tobacco quality is disease detection which needs to be performed before the smoking (oven) process during the manufacturing process. The tobacco leaf price is subject to the quality of the leaf; hence, factories need to do quality control throughout production to produce quality products. In most cases, the tobacco leaf picking procedure is done manually, resulting in uneven quality due to downtime and other circumstances. Reducing downtime requires adequate facilities and infrastructure in the manufacturing process. The solution to this problem is developing modern image processing technology. This study aims to detect disease in tobacco leaves, and the appropriate algorithm for selecting the best quality tobacco leaves with the help of image processing methods. Thus, this paper designs a Faster R-CNN detection method to address the problem. The evaluation results found that the highest value of the F1 score was 81.43%, which is the highest value obtained from training at the 8000th step. © 2023 IEEE.
State Polytechnic of Malang, Department of Electrical Engineering, Malang, Indonesia; State Polytechnic of Malang, Department of Information Technology, Malang, Indonesia