We propose a new approach for paragraph recognition in document images by
spatial graph convolutional networks (GCN) applied on OCR text boxes. Two
steps, namely line splitting and line clustering, are performed to extract
paragraphs from the lines in OCR results. Each step uses a beta-skeleton graph
constructed from bounding boxes, where the graph edges provide efficient
support for graph convolution operations. With only pure layout input features,
the GCN model size is 3~4 orders of magnitude smaller compared to R-CNN based
models, while achieving comparable or better accuracies on PubLayNet and other
datasets. Furthermore, the GCN models show good generalization from synthetic
training data to real-world images, and good adaptivity for variable document
styles.

本文提出了一种新的方法，使用空间图卷积网络（GCN）应用于 OCR 文本框，在文档图像中识别段落。通过线路分割和线路聚类两个步骤，从 OCR 结果的行中提取段落。每个步骤使用从边界框建立的贝塔骨架图，其中图边提供了有利于图卷积操作的高效支持。仅使用纯布局输入特征，GCN 模型大小比基于 R-CNN 的模型小 3~4 个数量级，同时在 PubLayNet 和其他数据集上实现可比较或更好的准确性。此外，GCN 模型在从合成训练数据到真实世界图像的转换，以及针对可变文档样式的适应性方面表现良好。