method: TPS-ResNet v12019-04-30

Authors: Jeonghun Baek, Moonbin Yim, Sungrae Park, and Hwalsuk Lee

Description: We used Thin-plate-spline (TPS) based Spatial transformer network (STN) which normalizes the input text images, ResNet based feature extractor, BiLSTM, and attention mechanism.
This model was developed based on the analysis of scene text recognition modules.
See our paper and source code.

[Training Data]
At first, we generated the Chinese synthetic datasets by MJSynth and SynthText code, then pre-trained our model with the synthetic dataset and real dataset (ArT, LSVT, ReCTS, and RCTW). After that, we finetuned it with ReCTS data.

method: resnet101lstm2019-04-27

Authors: 柳博方(Bofang Liu), 山东大学(Shandong University), 张锦华(Jinhua Zhang), 广东工业大学(Guangdong University of Technology)

Description: I regard the Text Line Recognition as a sequence question,so I adopted a convolutional neural network (e.g. ResNet) to extractor the image's feature and transform the features to sequence. Then I take a advantage of a recurrent neural work (e.g. LSTM) to recognition the Text Line.
My email is yingsunwangjian@gmail.com

method: baseline2019-04-30

Authors: zhujinyi(from Peking University)

Description: use CRNN to extract feature, deploy an attentional seq-to-seq model as the decoder
train da

Ranking Table

Description Paper Source Code
DateMethodResult
2019-04-30TPS-ResNet v194.77%
2019-04-27resnet101lstm78.26%
2019-04-30baseline75.50%

Ranking Graphic