Skip to yearly menu bar Skip to main content


Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data

Xinghua Lou · Ken Kansky · Wolfgang Lehrach · CC Laan · Bhaskara Marthi · D. Phoenix · Dileep George

Area 5+6+7+8 #86

Keywords: [ Structured Prediction ] [ Graphical Models ] [ (Other) Applications ] [ (Other) Probabilistic Models and Methods ] [ (Application) Object and Pattern Recognition ] [ (Application) Computer Vision ]


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.

Live content is unavailable. Log in and register to view live content