A paper published at BMVC2023

A paper published at BMVC2023

Created
September 1, 2023
Tags
PaperComputer Vision
Updated
November 20, 2023

We are excited to announce that our paper “Selective scene text removal” has been accepted to BMVC2023.

Hayato Mitani, Akisato Kimura, Seiichi Uchida, “Selective scene text removal,” British Machine Vision Conference (BMVC), 2023.
image

Scene text removal (STR) is the image transformation task to remove text regions in scene images (see the center of the above figure). The conventional STR methods remove all scene text. This means that the existing methods cannot select text to be removed.

In this paper, we propose a novel task setting named selective scene text removal (SSTR) that removes only target words specified by the user (see the right of the above figure). Although SSTR is a more complex task than STR, the proposed multi-module structure enables efficient training for SSTR.

Details can be seen at the official proceeding https://proceedings.bmvc2023.org/521/ or arxiv preprint https://arxiv.org/abs/2309.00410 , and a sample code can be found at GitHub https://github.com/mitanihayato/Selective-Scene-Text-Removal .