Update app.py

app.py

@@ -40,8 +40,14 @@ BASE_LOCATION = [0, 23]
 RULES = """<h1>OSV-5M (plonk)</h1>
 <center><img width="256" alt="Rotating globe" src="https://upload.wikimedia.org/wikipedia/commons/6/6b/Rotating_globe.gif"></center>
 <h2> Instructions </h2>
-<h3> Click on the map 🗺️ (left) to the location at which you think the image 🖼️ (right) was captured! </h3>
-<h3> Click "Select" to finalize your selection and then "Next" to move to the next image. </h3>
+<h3> Click on the map 🗺️ (left) to the location at which you think the image 🖼️ (right) was captured!</h3>
+<h3> Click "Select" to finalize your selection and then "Next" to move to the next image.</h3>
+
+<h2> AI Competitors </h2>
+<h3> You will compete against two AIs: Plonk-AI (our best model) and Baseline-AI (a simpler approach). These AIs have not been trained on any of the images you will see; in fact, they haven't seen anything within a 1km radius of them. Like you, the AIs will need to pick up on geographic clues to pinpoint the locations of the images.</h3>
+
+<h2> Geoscore </h2>
+<h3> The geoscore is calculated based on how close each guess is to the true location as in Geoguessr, with a maximum of 5000 points,  using the formula := 5000*exp(-d/1492.7) </h3>
 """
 css = """
 @font-face {
@@ -274,6 +280,10 @@ class Engine(object):
     def load_images_and_coordinates(self, csv_file):
         # Load the CSV
         df = pd.read_csv(csv_file)
+        # Put image with id 732681614433401 on the top and then all the rest below
+        df['id'] = df['id'].astype(str)
+        df = pd.concat([df[df['id'] == '495204901603170'], df[df['id'] != '495204901603170']])
+        df = pd.concat([df[df['id'] == '732681614433401'], df[df['id'] != '732681614433401']])
 
         # Get the image filenames and their coordinates
         self.images = [os.path.join(self.image_folder, f"{img_path}.jpg") for img_path in df['id'].tolist()[:]]