Upload README.md

README.md

@@ -23,7 +23,7 @@ Official repo for paper ["Bridge Diffusion Model: bridge non-English language-na
 
 ## Method 
 BDM entails the utilization of a backbone-branch network architecture akin to ControlNet[[7]](#7), model structure illustrated in the following    
-<p align="center"><img src="docs\BDM_structure.png" alt= “BDM” width="400" height="300"></p>      
+<p align="center"><img src="BDM_structure.png" alt= “BDM” width="400" height="300"></p>      
 <p align="center">Fig.1 BDM model structure</p> 
 
 The backbone part serves as a good diffusion initialization and will be frozen during training, which could be from any pretrained diffusion TTI model. We leverage Stable Diffusion 1.5 in current implementation. The branch part servers as language-native semantics injection module, whose parameters will be trained with language-native text-image pairs. 
@@ -33,9 +33,9 @@ For model inference, language-native positive prompts as well as negative ones w
 
 ## Evaluation
 Here are several image generation illustrations for our BDM, with Chinese-native TTI capability and integrated with different English TTI communty techniques.
-<p align="center"><img src="docs\Chinese_concepts.png" alt= “Chinese_concepts” width="600" height="550"></p>    
+<p align="center"><img src="Chinese_concepts.png" alt= “Chinese_concepts” width="600" height="550"></p>    
 <p align="center">Fig.2 Chinese unique concepts</p>
-<p align="center"><img src="docs\different_base_model.png" alt= “different_base_model” width="600" height="650"></p>  
+<p align="center"><img src="different_base_model.png" alt= “different_base_model” width="600" height="650"></p>  
 <p align="center">Fig.3 Different English branch</p>     
 
 For more illustrations and details, please refer to our paper ["Bridge Diffusion Model: bridge non-English language-native text-to-image diffusion model with English communities"](https://arxiv.org/abs/2309.00952)