Table Detection for Visually Rich Document Images
This work addresses table detection for document understanding, offering incremental improvements to existing methods.
The paper tackles the problem of table detection in visually rich documents by proposing a decoupled IoU loss to measure and reduce information loss, and improves SparseR-CNN with Gaussian noise augmentation and many-to-one label assignments, achieving consistent state-of-the-art performance across multiple datasets.
Table Detection (TD) is a fundamental task to enable visually rich document understanding, which requires the model to extract information without information loss. However, popular Intersection over Union (IoU) based evaluation metrics and IoU-based loss functions for the detection models cannot directly represent the degree of information loss for the prediction results. Therefore, we propose to decouple IoU into a ground truth coverage term and a prediction coverage term, in which the former can be used to measure the information loss of the prediction results. Besides, considering the sparse distribution of tables in document images, we use SparseR-CNN as the base model and further improve the model by using Gaussian Noise Augmented Image Size region proposals and many-to-one label assignments. Results under comprehensive experiments show that the proposed method can consistently outperform state-of-the-art methods with different IoU-based metrics under various datasets and demonstrate that the proposed decoupled IoU loss can enable the model to alleviate information loss.