Year: 2025
Type: article
Abstract: Tea drinking has been a part of culture for thousands of years plays a crucial role in daily life in today's world.It is difficult to find a person who does not drink tea 2-3 times a day, especially in South Asia.So, this important drink in our daily lives should be fresh and risk-free.It is very important and significant to detect the tea leaf disease timely in order to more production of fresh and risk-free tea.To detect tea leaf disease, numerous conventional methods are established including CNN, Deep CNN, DNN, AX-Retina Net, improved DCNN, YOLOv5, YOLOv7, and multi-object image segmentation etc.This research proposed a solution of hybrid model using YOLOv7 and DCNN for tea leaf disease detection with an improved accuracy rate.Here, the strong object detection features of YOLOv7 were utilized for identifying the detected regions and the deep learning methodology of DCNN was employed for identifying disease type accurately.This proposed model was experimented with a customized dataset of approximate 5,000 image that is collected from Bangladesh tea research institute (BTRI) and forest department, Bangladesh (FDB) achieved a mAP of 97.6%, a Precision of 95.2%, and an Accuracy of 97.8% for detecting the tea leaf disease.
Authors: Md Zahidul Kabir, Md Sourav Hossen, Sumiya Kaisar Keya
Institution: Gopalganj Science and Technology University
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4409840853