Font Generation with Missing Impression Labels

Our goal is to generate fonts with specific impressions, by training a generative adversarial network with a font dataset with impression labels. The main difficulty is that font impression is ambiguous and the absence of an impression label does not always mean that the font does not have the impression. This paper proposes a font generation model that is robust against missing impression labels. The key ideas of the proposed method are (1) a co-occurrence-based missing label estimator and (2) an impression label space compressor. The first is to interpolate missing impression labels based on the co-occurrence of labels in the dataset and use them for training the model as completed label conditions. The second is an encoder-decoder module to compress the high-dimensional impression space into low-dimensional. We proved that the proposed model generates high-quality font images using multi-label data with missing labels through qualitative and quantitative evaluations. Our code is available at https://github.com/SeiyaMatsuda/Font-Generation-with-Missing-Impression-Labels.

I. Introduction

Impressions of fonts enrich typographic designs, but they are subjective and often ambiguous. Fig. 1 shows three fonts and their impression labels from the MyFonts dataset [1]. The impression labels are attached by crowdsourcing; various font experts and non-experts freely attach the labels to each font. Moreover, the impression labels are open-vocabulary; there is no pre-defined list of impression labels. Consequently, impression labels attached are often incomplete. For example, abdominal-krunch in Fig. 1 could have the impression labels thick and bold. On the other hand, it is too optimistic to expect the complete impression labels, by considering the ambiguity in impressions. In other words, it is difficult to determine unanimously whether a certain font has a certain impression.