Context Synthetic digits with noisy backgrounds for testing robustness of classification algorithms. ---------- Content This dataset contains 12,000 synthetically generated images of English digits embedded on random backgrounds. The images are generated with varying fonts, colors, scales and rotations. The backgrounds are randomly selected from a [***subset***][3] of [***COCO***][4] dataset. ---------- Acknowledgements @article{roy2018effects,
title={Effects of Degradations on Deep Neural Network Architectures},
author={Roy, Prasun and Ghosh, Subhankar and Bhattacharya, Saumik and Pal, Umapada},
journal={arXiv preprint arXiv:1807.10108},
year={2018}
} This dataset is created as a benchmark dataset for the work on [***Effects of Degradations on Deep Neural Network Architectures***][1].
The source code is publicly available on [***GitHub***][2]. [1]: https://arxiv.org/abs/1807.10108 [2]: https://github.com/prasunroy/cnn-on-degraded-images [3]: http://images.cocodataset.org/zips/val2017.zip [4]: http://cocodataset.org