Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data
Abstract
We demonstrate that a generative model for object shapes can achieve state of the art results on challenging scene text recognition tasks, and with orders of magnitude fewer training images than required for competing discriminative methods. In addition to transcribing text from challenging images, our method performs fine-grained instance segmentation of characters. We show that our model is more robust to both affine transformations and non-affine deformations compared to previous approaches.
Cite
Text
Lou et al. "Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data." Neural Information Processing Systems, 2016.Markdown
[Lou et al. "Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data." Neural Information Processing Systems, 2016.](https://mlanthology.org/neurips/2016/lou2016neurips-generative/)BibTeX
@inproceedings{lou2016neurips-generative,
title = {{Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data}},
author = {Lou, Xinghua and Kansky, Ken and Lehrach, Wolfgang and Laan, Cc and Marthi, Bhaskara and Phoenix, D. and George, Dileep},
booktitle = {Neural Information Processing Systems},
year = {2016},
pages = {2793-2801},
url = {https://mlanthology.org/neurips/2016/lou2016neurips-generative/}
}