HENet: Forcing a Network to Think More for Font Recognition

TL;DR

This paper introduces HENet, a font recognition model that employs a pluggable HE Block to compel the network to consider more complex features, improving accuracy without additional inference interactions.

Contribution

The novel HE Block module forces the network to utilize less obvious features, enhancing font recognition performance on challenging similar fonts.

Findings

HENet outperforms existing font recognition systems.

The HE Block improves recognition of hard font examples.

HENet achieves high accuracy on character and word-level datasets.

Abstract

Although lots of progress were made in Text Recognition/OCR in recent years, the task of font recognition is remaining challenging. The main challenge lies in the subtle difference between these similar fonts, which is hard to distinguish. This paper proposes a novel font recognizer with a pluggable module solving the font recognition task. The pluggable module hides the most discriminative accessible features and forces the network to consider other complicated features to solve the hard examples of similar fonts, called HE Block. Compared with the available public font recognition systems, our proposed method does not require any interactions at the inference stage. Extensive experiments demonstrate that HENet achieves encouraging performance, including on character-level dataset Explor_all and word-level dataset AdobeVFR