Deep learning is used for coffee disease classification

The largest coffee producer and exporter in the world is Brazil. This study by João Vitor Yukio Bordin Yamashita, and João Paulo R.R. Leite identifies coffee diseases on lea es on low cost coffee farms in loco without internet connection. In order to achieve production quality and productivity then the coffee diseases on the leaves needs to be detected early. In order to detect the diseases two data were used and were in the form of images with over 600 images used. The results showed that the diseases were mostly from coffee farms that are not exposed to a lot of resources.The study mentioned that convolutional neural networks to detect the diseases on coffee leaves was verified and does not require an internet connection. This method can benefit low cost coffee producers that have less resources. The models showed the following results, “almost all performance indices were increased in the cascade model in relation to those from the single-stage model, except for those related to Rust class, which had a small decrease. The single- stage model showed difficulties in diseases with more subtle or similar characteristics, such as Cercospora, Phoma and Leaf Miner.” The results show that the datasets need to be improved as well as procedure of the onboard system. It has been mentioned that some coffee plants might be affected by more than one disease. More research needs to be out taken in this sector.

dayal.prina