Jute Plant Disease Detection by obtaining RoI with SGM using Image Processing Technique and CNN

In today's technological era, the demand for high-quality fiber is increasing as its usage grows. Proportionally jute disease prevention has become an urgent topic to fulfill the demand for qualitative fiber. Training a machine learning model to predict these diseases requires a significant amount of time to manually annotate a dataset. This study aims to identify the Region of Interest (ROI) in an image containing an object. We propose a method, Self-Generated Mask (SGM), to extract an image's Region of Interest. Jute leaves and stems have been selected to obtain the RoI of objects contrasting in shape. As per our knowledge, no other method has been proposed to obtain the RoI of objects with dissimilar shapes. Using the dataset of jute disease containing 1772 images in 10 classes, our method has successfully extracted the RoI of jute stems and leaves. The CNN model training with the dataset generated by the SGM method has been performed close to the human-annotated dataset. The study reveals that the manual annotation time period can be reduced in a mentionable amount of time.