# Coloring With Limited Data: Few-Shot Colorization via Memory-Augmented Networks

## TL;DR

This paper introduces MemoPainter, a memory-augmented neural network that achieves high-quality image colorization with limited data, especially excelling in few-shot and one-shot scenarios, by capturing rare instances and using a novel unsupervised training loss.

## Contribution

The paper proposes MemoPainter, a novel memory-augmented colorization model with a threshold triplet loss for unsupervised training, enabling effective few-shot and one-shot colorization.

## Key findings

- Outperforms existing models in few-shot colorization quality
- Successfully captures rare instances during colorization
- Operates effectively without class labels using unsupervised training

## Abstract

Despite recent advancements in deep learning-based automatic colorization, they are still limited when it comes to few-shot learning. Existing models require a significant amount of training data. To tackle this issue, we present a novel memory-augmented colorization model MemoPainter that can produce high-quality colorization with limited data. In particular, our model is able to capture rare instances and successfully colorize them. We also propose a novel threshold triplet loss that enables unsupervised training of memory networks without the need of class labels. Experiments show that our model has superior quality in both few-shot and one-shot colorization tasks.