diff --git a/doc/src/modle_arcitecture.md b/doc/src/modle_arcitecture.md index 47b75f0d..67c1434a 100644 --- a/doc/src/modle_arcitecture.md +++ b/doc/src/modle_arcitecture.md @@ -1,17 +1,25 @@ # Model Arcitecture The implemented arcitecure of Deepspeech2 online model is based on [Deepspeech2 model](https://arxiv.org/pdf/1512.02595.pdf) with some changes. - The figure of arcitecture is shown in ![image](../image/ds2onlineModel.png). - The model is mainly composed of 2D convolution subsampling layer and single direction rnn layers. To illustrate the model implementation in detail, 5 parts is introduced. + The model is mainly composed of 2D convolution subsampling layer and single direction rnn layers. + To illustrate the model implementation clearly, 5 parts is described in detail. 1. Feature Extraction. 2. 2D Convolution subsampling layer. 3. RNN layer with only forward direction. 4. Softmax Layer. 5. CTC Decoder. +The arcitecture of the model is shown in Fig.1. +
+
+
Fig.1 The Arcitecture of deepspeech2 online modle
+