'Four-panel Cartoon Image Dataset'

We have developed a new dataset called the "Four-panel Cartoon Image Dataset," which contains both grayscale and color images of four-panel cartoons spanning from 1920 to 2022. To collect the images, we obtained them from various online sources, including digitized newspapers such as Chosun Ilbo and Dong-a Ilbo. We used DarkLabel2.4 software to create bounding boxes for the four-panel images, utilizing two types of formats: 4x1 and 2x2. We generated labeling information for each image file in the You Only Look Once (YOLO) format and saved it in text files for object detection in deep learning. The dataset consists of a total of 161 image files in JPG format, which we randomly divided into training, validation, and testing sets using Roboflow software.