Bounding box guidance improves reading tables from images
Enhancing Table Structure Recognition via Bounding Box Guidance
Computer Vision and Pattern Recognition
Summary
Understanding tables in pictures requires figuring out where each cell is and how the table is organized. Current methods that turn table images into text codes often miss using the exact cell locations, causing mistakes especially in tricky images. The authors created a new method, BGTR, that first finds cell boxes and then uses that to help write the table layout correctly. They also made a new set of fake table pictures to help train the system better where real data is scarce. Tests show their method works better than previous ones on several table reading challenges.
Table structure recognitionBounding boxHTML sequenceImage-to-sequenceCell detectionSynthetic dataProgressive trainingNatural scene tablesBenchmark datasetsSOTA performance
Authors
Lei Hu, Shuangping Huang
Abstract
Table Structure Recognition (TSR) aims to extract the bounding boxes of cells and table structure (e.g., HTML) from table images. Although current approaches have made significant progress, the latest image-to-sequence methods overlook the explicit utilization of the bounding box information when predicting HTML sequences, leading to error predictions in complex scenes. In this paper, we introduce a novel framework BGTR (Bounding Box-Guided Table Recognizer). To more effectively utilize bounding box information, we first predict the bounding boxes of cells and then use this information to guide the generation of HTML sequences. While utilizing bounding box information can enhance the accuracy of HTML sequences, for natural scene tables, the data volume is too small to allow for sufficient training of bbox-guided HTML generation. In response, we adopt a progressive training method for natural scene tables and introduce SNSTab, a synthetically generated natural scene table dataset. Our experiments on five benchmark datasets demonstrate SOTA performance.