OmniPrint-NeurIPS-paper-experiments

This repository contains the code to reproduce the experiments described in the paper OmniPrint: A Configurable Printed Character Synthesizer, which is accepted at NeurIPS 2021 Track Datasets and Benchmarks.

The main repository of OmniPrint is https://github.com/SunHaozhe/OmniPrint

Instructions

git clone https://github.com/SunHaozhe/OmniPrint
git clone https://github.com/SunHaozhe/OmniPrint-NeurIPS-paper-experiments

Copy and paste each subfolder of OmniPrint-NeurIPS-paper-experiments/experiments/ into OmniPrint/.

• OmniPrint-NeurIPS-paper-experiments/experiments/baseline_algorithms: Section 4.1 Few-shot learning
• OmniPrint-NeurIPS-paper-experiments/experiments/baseline_algorithms_vary_train_size: Section 4.3 Influence of the number of meta-training episodes for few-shot learning
• OmniPrint-NeurIPS-paper-experiments/experiments/baseline_algorithms_Z: Section 4.2 Other meta-learning paradigms
• OmniPrint-NeurIPS-paper-experiments/experiments/regression: Section 4.5 Character image regression tasks
• OmniPrint-NeurIPS-paper-experiments/experiments/transfer: Section 4.4 Domain adaptation
• OmniPrint-NeurIPS-paper-experiments/experiments/transfer_mnist: Section 4.4 Domain adaptation

For the slurm scripts, some sbatch parameters have been removed (cluster partition, cluster QoS, which nodes to use). Please complete and adapt them according to your compute resources.

Citation

@inproceedings{sun2021omniprint,
title={OmniPrint: A Configurable Printed Character Synthesizer},
author={Haozhe Sun and Wei-Wei Tu and Isabelle M Guyon},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1)},
year={2021},
url={https://openreview.net/forum?id=R07XwJPmgpl}
}