Attention-based Fully Gated CNN-BGRU for Russian Handwritten Text

摘要: This research approaches the task of handwritten text with attention encoder-decoder networks that are trained on Kazakh and Russian language. We developed a novel deep neural network model based Fully Gated CNN, supported by Multiple bidirectional GRU Attention mechanisms to manipulate sophisticated features achieve 0.045 Character Error Rate (CER), 0.192 Word (WER) 0.253 Sequence (SER) for first test dataset 0.064 CER, 0.24 WER 0.361 SER second dataset. Also, we propose fully gated layers taking advantage multiple output feature from Tahn input feature, this proposed work achieves better results experimented our Handwritten & Database (HKR). Our is HKR demonstrates state-of-the-art most other existing models.