OmniPrint: A Configurable Printed Character Synthesizer
arXiv:2201.06648
Abstract
We introduce OmniPrint, a synthetic data generator of isolated printed characters, geared toward machine learning research. It draws inspiration from famous datasets such as MNIST, SVHN and Omniglot, but offers the capability of generating a wide variety of printed characters from various languages, fonts and styles, with customized distortions. We include 935 fonts from 27 scripts and many types of distortions. As a proof of concept, we show various use cases, including an example of meta-learning dataset designed for the upcoming MetaDL NeurIPS 2021 competition. OmniPrint is available at https://github.com/SunHaozhe/OmniPrint.
arXiv commentAccepted at 35th Conference on Neural Information Processing Systems (NeurIPS 2021) Track on Datasets and Benchmarks. https://openreview.net/forum?id=R07XwJPmgpl
References in corpus (17)
- Adam: A Method for Stochastic Optimization
- scikit-image: Image processing in Python
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Prototypical Networks for Few-shot Learning
- An overview of gradient descent optimization algorithms
- Learning Transferable Features with Deep Adaptation Networks
- Unsupervised Domain Adaptation by Backpropagation
- Deep Domain Confusion: Maximizing for Domain Invariance
- Unsupervised Learning of Video Representations using LSTMs
- Deep Subdomain Adaptation Network for Image Classification
- Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition
- Generalized Inner Loop Meta-Learning
- Cold Case: The Lost MNIST Digits
- MNIST-C: A Robustness Benchmark for Computer Vision
- UnrealText: Synthesizing Realistic Scene Text Images from the Unreal World
- Text Recognition in the Wild: A Survey
- Editing Text in the Wild