A Survey on Object Detection from Scientific Plots
John Mathew, Janaki Meena M · 2021
Charts have always been the most effective way to provide a concise and comprehensive method to represent statistical data. Several studies have been undertaken in the field of object detection from scientific plots, which concentrate on the process of plot element detection, extraction of the value depicted by the element, and analysis of the extracted values, to reconstruct the dataset table that was used to generate the plot. Major computer vision problems may be solved using the high accuracy object detection procedures attempted while performing data extraction from chart plots. Applications that benefit from chart mining include tabular data recreation, text to speech solutions for the visually impaired, as well as digitization of print research. This paper aims to survey the most effective existing methodologies for data extraction from charts, while attempting to develop a pipeline that encompasses key modules from the analyzed documents to provide a higher accuracy during object detection.