Table Structure Detection Based on PPLCNet Network
Linfeng Wu, Chenyu Wang, Yu Wu, Shunli Chen, Yiling Tan, Qiang Wang · 2024
This paper presents table structure recognition models based on PPLCNet, MobileNetV3, and ResNet, evaluated on the PubTabNet dataset. By incorporating residual structures and the CBAM attention mechanism into PPLCNet, we achieve an accuracy of 73.53% and a TEDS of 94.57%. Ablation and comparative experiments validate the effectiveness of these enhancements.