Update README.md


README.md

@@ -1,9 +1,12 @@
 ---
 license: afl-3.0
 ---
-This project page is about the pytorch code implementation of GlyphBERT by the HITsz-TMG research group.
+This project page is about the pytorch code implementation of GlyphBERT by the HITsz-TMG research group.  
 
-GlyphBERT is a Chinese pre-training model that includes Chinese character glyph features.It renders the input characters into images and designs them in the form of multi-channel location feature maps, and designs a two-layer residual convolutional neural network module to extract the image features of the characters for training.
+
+GlyphBERT is a Chinese pre-training model that includes Chinese character glyph features.It renders the input characters into images and designs them in the form of multi-channel location feature maps, and designs a two-layer residual convolutional neural network module to extract the image features of the characters for training.  
+
+![img.png](https://s3.amazonaws.com/moonup/production/uploads/1661697350102-621a2b96100edd793f521ab6.png)    
 
 The experimental results show that the performance of the pre-training model can be well improved by fusing the features of Chinese glyphs. GlyphBERT is much better than BERT in multiple downstream tasks, and has strong transferability.