Abstract:
Tea (Camellia sinensis) is one of the most economically important plantation
crops in Sri Lanka; however, foliar diseases such as blister blight and brown
blight continue to cause significant yield and quality losses if not identified at
early stages. Conventional disease detection methods rely heavily on manual
visual inspection, which is time-consuming, subjective, and difficult to scale
across large plantation areas. To address these limitations, this study proposes
a deep learning–based tea leaf disease detection system using image
recognition and a custom Convolutional Neural Network (CNN), with an
emphasis on real-time field applicability. A dataset of 1,200 tea leaf images was
collected from tea estates in the upcountry region of Sri Lanka under natural
field conditions. The dataset comprises three classes: healthy leaves, blister
blight–infected leaves, and brown blight–infected leaves. Images were
preprocessed using resizing, normalization, and data augmentation techniques
to enhance robustness against lighting variations and background noise. The
CNN model was trained and evaluated using a 70:15:15 train–validation–test
split. Experimental evaluation demonstrates that the proposed model achieves
an overall classification accuracy of 92.4% and an F1-score of 0.92, indicating
effective discrimination among disease classes. The trained model was
successfully integrated into an Android-based mobile application, enabling ondevice
inference with an average prediction time of less than two seconds per
image, making it suitable for real-time agricultural use without continuous
internet connectivity. Despite promising results, the study is limited by the
moderate dataset size and variability in lighting conditions during image
acquisition. Future work will focus on dataset expansion, improved robustness
under diverse field environments, and large-scale field validation. The
proposed system highlights the potential of deep learning–enabled mobile
solutions to support precision agriculture and early disease management in Sri
Lanka’s tea industry.
