Realistic text replacement with non-uniform style conditioning

TL;DR

This paper introduces a novel non-uniform style conditioning layer for realistic text replacement in images, enabling high-quality, indistinguishable edits without post-processing, and demonstrates superior performance on the ICDAR MLT dataset.

Contribution

The paper presents a new non-uniform style conditioning method integrated into an encoder-decoder ResNet architecture for realistic text replacement.

Findings

Outperforms existing methods on ICDAR MLT dataset

Achieves realistic, indistinguishable text replacement

Single-stage model with no post-processing

Abstract

In this work, we study the possibility of realistic text replacement, the goal of which is to replace text present in the image with user-supplied text. The replacement should be performed in a way that will not allow distinguishing the resulting image from the original one. We achieve this goal by developing a novel non-uniform style conditioning layer and apply it to an encoder-decoder ResNet based architecture. The resulting model is a single-stage model, with no post-processing. The proposed model achieves realistic text replacement and outperforms existing approaches on ICDAR MLT.