MTRNet++: One-stage Mask-based Scene Text Eraser

TL;DR

MTRNet++ is a one-stage, mask-based scene text eraser that achieves state-of-the-art results with a novel multi-branch architecture, attention mechanisms, and controllability, without relying on external masks.

Contribution

The paper introduces MTRNet++, a new one-stage text removal network with a multi-branch architecture and attention blocks, improving controllability and interpretability over prior methods.

Findings

Achieves state-of-the-art results on Oxford and SCUT datasets.

Effective multi-branch architecture with attention blocks demonstrated.

Operates with or without external masks, showing flexibility.

Abstract

A precise, controllable, interpretable and easily trainable text removal approach is necessary for both user-specific and large-scale text removal applications. To achieve this, we propose a one-stage mask-based text inpainting network, MTRNet++. It has a novel architecture that includes mask-refine, coarse-inpainting and fine-inpainting branches, and attention blocks. With this architecture, MTRNet++ can remove text either with or without an external mask. It achieves state-of-the-art results on both the Oxford and SCUT datasets without using external ground-truth masks. The results of ablation studies demonstrate that the proposed multi-branch architecture with attention blocks is effective and essential. It also demonstrates controllability and interpretability.