Boosting Modern and Historical Handwritten Text Recognition with Deformable Convolutions
This work addresses the problem of digitizing handwritten documents for archivists and researchers, but it is incremental as it builds on existing CNN-RNN approaches with a specific architectural modification.
The paper tackled the challenge of Handwritten Text Recognition (HTR) in free-layout pages, especially for historical documents, by proposing deformable convolutions to adapt to geometric variations in text, resulting in improved performance on modern and historical datasets.
Handwritten Text Recognition (HTR) in free-layout pages is a challenging image understanding task that can provide a relevant boost to the digitization of handwritten documents and reuse of their content. The task becomes even more challenging when dealing with historical documents due to the variability of the writing style and degradation of the page quality. State-of-the-art HTR approaches typically couple recurrent structures for sequence modeling with Convolutional Neural Networks for visual feature extraction. Since convolutional kernels are defined on fixed grids and focus on all input pixels independently while moving over the input image, this strategy disregards the fact that handwritten characters can vary in shape, scale, and orientation even within the same document and that the ink pixels are more relevant than the background ones. To cope with these specific HTR difficulties, we propose to adopt deformable convolutions, which can deform depending on the input at hand and better adapt to the geometric variations of the text. We design two deformable architectures and conduct extensive experiments on both modern and historical datasets. Experimental results confirm the suitability of deformable convolutions for the HTR task.