diff --git a/README.md b/README.md
index 8d623c6..400a24f 100644
--- a/README.md
+++ b/README.md
@@ -1,10 +1,15 @@
-<img src="images/logo.jpg"/>
+<p align="center">
+  <img src="images/logo.jpg"/>
+</p>
 
 # LAMDA: Label Matching Deep Domain Adaptation
 
 This is the implementation of paper **[LAMDA: Label Matching Deep Domain Adaptation](http://proceedings.mlr.press/v139/le21a/le21a.pdf)** which has been accepted at ICML 2021.
 
-<img src="images/Figure1.png" style="zoom:20%;" />
+<p align="center">
+  <img src="images/Figure1.png" width="400"/>
+</p>
+
 
 ## A. Setup
 
@@ -121,4 +126,4 @@ python run_lamda.py 1 webcam amazon format csv num_iters 20000 summary_freq 400
 
 ### D.2. GitHub repositories: 
 
-- Some parts of our code (e.g., VAT, evaluation, …) are rewritten with modifications from [DIRT-T](https://github.com/RuiShu/dirt-t).
\ No newline at end of file
+- Some parts of our code (e.g., VAT, evaluation, …) are rewritten with modifications from [DIRT-T](https://github.com/RuiShu/dirt-t).