From Pages to Panels: Deep Learning-based Comics Segmentation with CNN+RF

This research study employs a dual approach that involves RF and CNN to present a new means of detecting comic panels to access the education. The main objective therefore is to enhance the panel recognition and the precision of the same to aid in the preservation, searching as well and retrieval of the comics. A CNN is applied to extract features underlying the image and an RF model is applied under the classification step. From the results of our tests, we observe significant increases compared to baseline methods, the best of which is the suggested CNN+ RF with a 92.5% accuracy rate. Through this method, the capacity increases for organizing digital comic collections deftly as it furnishes a good response concerning the contingency and challenge of comic panels.