one more round of dataset format updates

dataset.py

@@ -1,5 +1,4 @@
 
-import os
 import datasets
 
 class ForestSegmentationDataset(datasets.GeneratorBasedBuilder):
@@ -7,36 +6,37 @@ class ForestSegmentationDataset(datasets.GeneratorBasedBuilder):
         return datasets.DatasetInfo(
             features=datasets.Features({
                 "sample_id": datasets.Value("string"),
-                "image": datasets.Image(),  # Path to all_bands.tif
-                "mask": datasets.Image(),   # Path to mask.tif
+                "image_paths": datasets.Sequence(datasets.Value("string")),  # list of GeoTIFF paths
+                "mask": datasets.Value("string"),  # path to mask
+                "start_timestamps": datasets.Sequence(datasets.Value("string")),
+                "end_timestamps": datasets.Sequence(datasets.Value("string")),
+                "sentinel_tile_ids": datasets.Sequence(datasets.Value("string")),
+                "bboxes": {
+                    "minx": datasets.Sequence(datasets.Value("float32")),
+                    "maxx": datasets.Sequence(datasets.Value("float32")),
+                    "miny": datasets.Sequence(datasets.Value("float32")),
+                    "maxy": datasets.Sequence(datasets.Value("float32")),
+                }
             }),
         )
 
     def _split_generators(self, dl_manager):
-        # Streaming mode: use iter_files to lazily iterate over image files
-        image_paths = dl_manager.iter_files("images")
+        sample_stream = dl_manager.iter_jsonl("index.jsonl")
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
-                gen_kwargs={"image_paths": image_paths}
+                gen_kwargs={"samples": sample_stream}
             )
         ]
 
-    def _generate_examples(self, image_paths):
-        for image_path in image_paths:
-            if not image_path.endswith("all_bands.tif"):
-                continue
-
-            # Extract sample_id from path: images/0000005/<timestamp>/all_bands.tif
-            parts = image_path.split("/")
-            if len(parts) < 3:
-                continue  # Skip malformed paths
-
-            sample_id = parts[-3]  # "0000005"
-            mask_path = f"masks/{sample_id}/mask.tif"
-
-            yield sample_id, {
-                "sample_id": sample_id,
-                "image": image_path,
-                "mask": mask_path,
+    def _generate_examples(self, samples):
+        for sample in samples:
+            yield sample["sample_id"], {
+                "sample_id": sample["sample_id"],
+                "image_paths": sample["image_paths"],
+                "mask": sample["mask_path"],
+                "start_timestamps": sample["start_timestamps"],
+                "end_timestamps": sample["end_timestamps"],
+                "sentinel_tile_ids": sample["sentinel_tile_ids"],
+                "bboxes": sample["bboxes"],
             }