Optimization-Based Algorithm for Detecting Key Information in Yolov7 Literature

Literature is a reservoir of knowledge, which contains the research results, theoretical discussions and practical experiences of previous generations. Literature key information including title, author, journal name, keywords, DOI and publisher is an important part of scientific and technological literature, and it is also the key basis for scientific and technological literature e-resources repositories and researchers to store, manage and retrieve literature. In order to accurately extract the information, an optimization-based YOLOv7 literature key information detection algorithm is proposed. And the three optimization measures include the introduction of CSPResNet, CSPPAN and SA attention mechanism. The experimental results show that the optimized YOLOv7-based literature key information detection algorithm achieves an average accuracy (AP0.5) of 94.0% and takes 8.99 frames/second, which maintains a faster detection speed while having a better accuracy.