Update README.md

README.md

@@ -54,8 +54,7 @@ model-index:
     - type: city accuracy
       value: 5.9
 ---
-
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/654bb2591a9e65ef2598d8c4/0Z-GMa6SSLgXFmrplC0WD.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/654bb2591a9e65ef2598d8c4/mmTZy5ELTwLiLap8pO4xV.png)
 
 # OpenStreetView-5M <br><sub>The Many Roads to Global Visual Geolocation 📍🌍</sub>
 
@@ -71,7 +70,7 @@ Our dataset was used in an extensive benchmark of which we provide the best mode
 For more details and results, please check out our [paper](arxiv) and [project page](https://imagine.enpc.fr/~guillaume-astruc/osv-5m).  
 
 ### Inference 🔥
-![image/png](https://cdn-uploads.huggingface.co/production/uploads/654bb2591a9e65ef2598d8c4/mmTZy5ELTwLiLap8pO4xV.png)
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/654bb2591a9e65ef2598d8c4/0Z-GMa6SSLgXFmrplC0WD.png)
 
 Our best model on OSV-5M can also be found on [huggingface](https://huggingface.co/osv5m/baseline).  
 First download the repo `!git clone https://github.com/gastruc/osv5m`.