You-Only-Look-Once-v5 Based Table Detection for Academic Papers

Tables are an important module for summarizing research results in academic papers. Quickly obtaining table information can improve the efficiency of scientific researchers in analyzing experimental data. This paper proposes a table detection method for academic paper based on YOLOv5 (You Only Look Once v5). First, the academic paper is preprocessed to fit the detection model, and then a YOLOv5 based neural network is used to detect tables quickly. Experiment results show that the detection accuracy and speed of the proposed approach are superior to the competitive approaches.