Learning Multimodal Affinities for Textual Editing in Images

TL;DR

This paper introduces an unsupervised deep learning method to learn multimodal affinities among textual elements in document images, enabling semantic clustering and editing of text content, style, and layout.

Contribution

It presents a novel unsupervised deep optimization approach for learning affinities between textual entities in document images considering multimodal information.

Findings

Effective clustering of textual entities into semantic groups.

Applicable to diverse document images with varying layouts.

Supports editing operations on content, appearance, and geometry.

Abstract

Nowadays, as cameras are rapidly adopted in our daily routine, images of documents are becoming both abundant and prevalent. Unlike natural images that capture physical objects, document-images contain a significant amount of text with critical semantics and complicated layouts. In this work, we devise a generic unsupervised technique to learn multimodal affinities between textual entities in a document-image, considering their visual style, the content of their underlying text and their geometric context within the image. We then use these learned affinities to automatically cluster the textual entities in the image into different semantic groups. The core of our approach is a deep optimization scheme dedicated for an image provided by the user that detects and leverages reliable pairwise connections in the multimodal representation of the textual elements in order to properly learn the affinities. We show that our technique can operate on highly varying images spanning a wide range of documents and demonstrate its applicability for various editing operations manipulating the content, appearance and geometry of the image.