Abstract:
"Agriculture plays a pivotal role in Sri Lanka’s economy, notably through tea production, which is
a significant source of foreign exchange earnings. However, the sector is currently grappling with
a critical issue: the widespread occurrence of diseases affecting tea leaves. These diseases not only
diminish the quality of the tea but also lead to a decrease in overall production. The health of the
tea plants is compromised, impacting essential functions such as fertilization.
In response to this problem, innovative techniques involving image processing and Convolutional
Neural Networks (CNN) have been employed. The process begins with the validation of the image
dataset by professionals from the Tea Research Institute (TRI). Following this, the images undergo
a preprocessing phase, which includes data augmentation techniques to eliminate any background
distractions and to resize the images for optimal input into the network. This step is crucial as it
prepares the data for more effective analysis by the CNN.
The results from the model are promising, with an accuracy score of 88%. This indicates that the
model is highly effective in classifying images and identifying tea leaf diseases. However, there is
room for improvement. By expanding the dataset—adding more images for the model to learn
from—the accuracy and performance of the CNN-based method can be further improved. This
approach holds the potential to revolutionize the identification process for tea leaf diseases, making
it faster, more accurate, and less reliant on expert availability."
