Few-shot Knowledge Transfer for Fine-grained Cartoon Face Generation

TL;DR

This paper introduces a two-stage training method for generating fine-grained cartoon faces across multiple groups, effectively transferring knowledge from a data-rich group to groups with limited samples.

Contribution

It proposes a novel two-stage training process that creates group-specific branches to learn unique features with few samples, improving cartoon face generation quality.

Findings

Effective knowledge transfer to groups with limited data

High-quality cartoon face generation across multiple groups

Maintains style consistency while capturing group-specific features

Abstract

In this paper, we are interested in generating fine-grained cartoon faces for various groups. We assume that one of these groups consists of sufficient training data while the others only contain few samples. Although the cartoon faces of these groups share similar style, the appearances in various groups could still have some specific characteristics, which makes them differ from each other. A major challenge of this task is how to transfer knowledge among groups and learn group-specific characteristics with only few samples. In order to solve this problem, we propose a two-stage training process. First, a basic translation model for the basic group (which consists of sufficient data) is trained. Then, given new samples of other groups, we extend the basic model by creating group-specific branches for each new group. Group-specific branches are updated directly to capture specific appearances for each group while the remaining group-shared parameters are updated indirectly to maintain the distribution of intermediate feature space. In this manner, our approach is capable to generate high-quality cartoon faces for various groups.