Detecting and Removing Text in the Wild

Abstract

Scene text removal is a challenging task that aims to erase wild text regions that include text strokes and their ambiguous boundaries, such as embossing, shade, or flare. The challenging issues raised in the wild are not completely addressed by the existing methods. To address these issues, we propose a new loss function for blending two tasks in a new network structure that depicts wild text regions in a soft mask and selectively inpaints them into a sensible background. The proposed loss function aids the learning of two seemingly separate tasks in a synergistic way via the soft mask to achieve remarkable performance in scene text removal. We validate our method through qualitative and quantitative comparisons, and region-wise analysis, showing that our method outperforms existing methods.

E2f-Net: Eyes-to-Face Inpainting Via Stylegan Latent Space

“…Recently, learning-based techniques such as deep convolutional neural networks (CNNs) and generative adversarial networks (GANs) have been widely used for a variety of image inpainting tasks, such as eliminating objects [7,8], noises [9], texts [10], and masks [11]. Usually, the proposed CNN-based methods are classified into three categories including coarse-to-fine, coarse-andfine, and structural guidance-based methods.…”

Don't Forget Me: Accurate Background Recovery for Text Removal via Modeling Local-Global Context

“…MTRNet++ [39] shared the same spirit with EraseNet, but separately encoded the image content and text mask in two branches. Cho et al [6] proposed to jointly predict the text stroke and inpaint the background, allowing the model to focus only on the restoration of text stroke regions. Wang et al [48] presented PERT, which contained a novel progressive structure with shared parameters to remove text more thoroughly, and a region-based modification strategy to effectively guide the erasure process only on text regions.…”