Character Recognition via a Compact Convolutional Neural Network

Optical Character Recognition (OCR) in the scanned documents has been a well-studied problem in the past. However, when these characters come from the natural scenes, it becomes a much more challenging problem, as there exist many difficulties in these images, e.g., illumination variance, cluttered backgrounds, geometry distortion. In this paper, we propose to use a deep learning method that based on the convolutional neural networks to recognize this kind of characters and word in the scene images. Based on the original VGG-Net, we focus on how to make a compact architecture on this net, and get both the character and word recognition results under the same framework. We conducted several experiments on the benchmark datasets of the natural scene images. The experiments has shown that our method can achieve the state-of-art performance and at the same time has a more compact representation.