Fonts-2-Handwriting: A Seed-Augment-Train framework for universal digit classification
Vinay Uday Prabhu, Sanghyun Han, Dian Ang Yap, Mihail Douhaniaris, Preethi Seshadri、John Whaley UnifyID AI Labs Redwood City, CA 94063, USA
Abstract
In this paper, we propose a Seed-Augment-Train/Transfer (SAT) framework that contains a synthetic seed image dataset generation procedure for languages with different numeral systems using freely available open font file datasets. This seed dataset of images is then augmented to create a purely synthetic training dataset, which is in turn used to train a deep neural network and test on held-out real world handwritten digits dataset spanning five Indic scripts, Kannada, Tamil, Gujarati, Malayalam, and Devanagari. We showcase the efficacy of this approach both qualitatively, by training a Boundary-seeking GAN (BGAN) that generates realistic digit images in the five languages, and also quantitatively by testing a CNN trained on the synthetic data on the real-world datasets. This establishes not only an interesting nexus between the font-datasets-world and transfer learning but also provides a recipe for universal-digit classification in any script.
中文速览
针对印度语系（Indic scripts）手写数字识别长期缺乏大规模标注数据集的问题，研究者提出了一套名为"种子-增强-训练/迁移"（Seed-Augment-Train/Transfer，SAT）的框架：先从免费开源字体文件中渲染出数字"种子图像"，再依次经过弹性形变（浅层增强）和条件生成对抗网络 ACGAN（深层增强），生成大规模纯合成训练数据，最终用卷积神经网络在五种印度语言（卡纳达语、泰米尔语、古吉拉特语、马拉雅拉姆语、梵文天城体）的真实手写数字集上完成测试。纯合成数据训练的分类准确率在60%–75%之间，而仅混入约20%真实样本后准确率大幅提升，部分语言中个别数字的准确率从不足1%跃升至95%；此外，用合成数据训练的边界寻找 GAN（BGAN）也能生成视觉上逼真的手写数字图像。这项工作为任意文字体系的数字识别提供了一条低成本、可复现的通用路径，揭示了字体文件这一被忽视的半合成数据宝库与迁移学习之间的重要联系。
原文 arXiv:1905.08633;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1905.08633v1