fix

docs/en/model_zoo/recognizaton/attention_lstm.md → docs/en/model_zoo/recognition/attention_lstm.md

@@ -1,11 +1,11 @@
 RNN is usualy used for sequence data, it works well for videos’ time-step information of continues frames.  
  And it’s a basic popular method for video classification.  
- AttentionLstm use a bidirectional LSTM to encode the frames of video.   
+ AttentionLstm use a bidirectional LSTM to encode the frames of video.  
  Different from the traditional method where LSTM is in the output layer for the last time,  
  AttentionLstm increase a Attention layer, there is a adaptive weight for the hidden layer of each time,  
- and weighted to get the final feature vector.   
+ and weighted to get the final feature vector.  
  For Attention layer, refer to the paper: AttentionCluster.  
- 
+
 Data Prepare, refer to the page of dataset: (TODO)  
 Model training and valid, refer to the page of tutorial: (TODO)  
 The Hit@1 on Youtube-8M’ validation is 0.89, PERR is	0.8012, GAP is 0.8594.

docs/en/model_zoo/recognizaton/pp-tsm.md → docs/en/model_zoo/recognition/pp-tsm.md

@@ -1,8 +1,8 @@
-Based on paddle2.0, we improved TSM，almost not increase the number of model parameters and FLOPs,   
-to achieve the goal of improving the accuracy of detector as much as possible   
-while ensuring that the speed is almost unchanged.   
+Based on paddle2.0, we improved TSM，almost not increase the number of model parameters and FLOPs,  
+to achieve the goal of improving the accuracy of detector as much as possible  
+while ensuring that the speed is almost unchanged.  
 
-By combining multiple tricks, PP-TSM can achieve a better balance between effectiveness (73.5% Top1)   
+By combining multiple tricks, PP-TSM can achieve a better balance between effectiveness (73.5% Top1)  
 and efficiency (TODO), surpassing the existing 2D video models’s accuracy such as TSN and TSM  
 in multiple datasets such as UCF101、Kinetics-400。  
 

docs/en/model_zoo/recognizaton/tsm.md → docs/en/model_zoo/recognition/tsm.md

@@ -1,4 +1,4 @@
-TSM is Temporal Shift Module for Efficient Video Understanding, which is Accepted by ICCV 2019.   
+TSM is Temporal Shift Module for Efficient Video Understanding, which is Accepted by ICCV 2019.  
 TSM enjoys both high efficiency and high performance.  
 Specifically, it can achieve the performance of 3D CNN but maintain 2D CNN's complexity.  
 For detail , refer to the paper Temporal Shift Module for Efficient Video Understanding.