End-to-end Text Recognition Using Convolutional Neural Networks

摘要

1.Recognizing end-to-end text in natural images is still a challenging problem as traditional systems have relied on models that incorporate hand engineered features or models that need large amounts of prior knowledge.rn2.Unlike character recognition for scanned documents, recognizing text in unconstrained images is complicated by a wide range of variations in backgrounds, textures, fonts, and lighting conditions.