Update app.py

app.py

@@ -6,7 +6,73 @@ import pandas as pd
 import os
 from itertools import islice
 
-# ... (rest of your code)
+# Configuration
+chunk_size = 100  # Size of the chunks to shuffle
+label_file = "labels.csv"
+
+# Load the Satellogic dataset (streaming)
+dataset = ev.load_dataset("satellogic", streaming=True)
+data_iter = iter(dataset)
+shuffled_chunk = []  # Initialize an empty list to hold the current chunk
+chunk_iter = None  # Initialize the chunk iterator
+
+# Initialize or load labels DataFrame
+if os.path.exists(label_file):
+    labels_df = pd.read_csv(label_file)
+else:
+    labels_df = pd.DataFrame(columns=["image_id", "bounds", "rating", "google_maps_link"])
+
+def get_next_image():
+    global data_iter, labels_df, shuffled_chunk, chunk_iter
+
+    while True:
+        # If we don't have a current chunk or it's exhausted, get a new one
+        if not shuffled_chunk or chunk_iter is None:
+            chunk = list(islice(data_iter, chunk_size))
+            if not chunk:  # If the dataset is exhausted, reset the iterator
+                print("Dataset exhausted, resetting iterator.")
+                data_iter = iter(ev.load_dataset("satellogic", streaming=True))
+                chunk = list(islice(data_iter, chunk_size))
+                if not chunk:
+                    print("Still no data after reset.")
+                    return None, "Dataset exhausted", None, None
+
+            random.shuffle(chunk)
+            shuffled_chunk = chunk
+            chunk_iter = iter(shuffled_chunk)
+
+        try:
+            sample = next(chunk_iter)
+            sample = ev.item_to_images("satellogic", sample)
+            image = sample["rgb"][0]
+            metadata = sample["metadata"]
+            bounds = metadata["bounds"]
+            google_maps_link = utils.get_google_map_link(sample, "satellogic")
+            image_id = str(bounds)
+
+            if image_id not in labels_df["image_id"].values:
+                return image, image_id, bounds, google_maps_link
+        except StopIteration:
+            # Current chunk is exhausted, reset chunk variables to get a new one in the next iteration
+            shuffled_chunk = []
+            chunk_iter = None
+
+def rate_image(image_id, bounds, rating, google_maps_link):
+    global labels_df
+
+    new_row = pd.DataFrame(
+        {
+            "image_id": [image_id],
+            "bounds": [bounds],
+            "rating": [rating],
+            "google_maps_link": [google_maps_link],
+        }
+    )
+    labels_df = pd.concat([labels_df, new_row], ignore_index=True)
+    labels_df.to_csv(label_file, index=False)
+
+    next_image, next_image_id, next_bounds, next_google_maps_link = get_next_image()
+    return next_image, next_image_id, next_bounds, next_google_maps_link
 
 # Gradio interface (remove initial_outputs from launch)
 iface = gr.Interface(