method: VIL2019-04-16

Authors: Xiaohan Zhang

Description: To recognise the content, we connected it with the first task.In the first task, we have obtained the detected region and its reliable bounding box, to make a dataset which contains numerous images of detected character sequence is not difficult.For the purpose of recognising, giving the images a specific naming format and matching it with a dictionary is important.And we apply some efficient character sequence recognition algorithm to the existed neural network.