Palm Leaf Manuscript Classification Based on Degradation Level Using Image Analysis Technique

Palm leaf manuscripts (PLM) are a vital but under-appreciated source of knowledge on the teachings and traditions of Indian culture and thus must be preserved for the future generations. The proposed work aims to aid this preservation effort by introducing a novel U-Net based deep learning model for the classifying PLMs according to their level of degradation thus helping researchers prioritize and optimize their work. In addition, a dataset consisting of 400 pre-processed images o PLM have also been prepared as part of this project. We have tested and compared the results of two models, U-Net and ResNet-50 and got the maximum accuracies of 83.6 and 62.7.