Update README.md

README.md

@@ -1,3 +1,58 @@
----
-license: cc-by-4.0
----
+---
+# ... (license, tags, etc.)
+dataset_info:
+  features:
+    # Fields directly from your metadata.jsonl
+    - name: file_name # This is standard for image datasets
+      dtype: string
+    - name: image_id   # The ID you added
+      dtype: string   # Since we made it string
+    - name: height
+      dtype: int32    # Assuming it's int32, adjust if needed
+    - name: width
+      dtype: int32    # Assuming it's int32
+    - name: dataset_origin
+      dtype: string
+    - name: distance_category
+      dtype: string
+    - name: objects
+      sequence:
+        feature:
+          - name: id
+            dtype: string # Annotation ID, now string
+          - name: area
+            dtype: float32
+          - name: bbox
+            sequence: float32
+          - name: category_id
+            dtype: int64 # Or int32, should be integer
+    - name: categories
+      sequence:
+        feature:
+          - name: id
+            dtype: int64 # Category ID
+          - name: name
+            dtype: string
+          - name: supercategory
+            dtype: string
+    # The 'image' field is often implicitly handled or can be added
+    - name: image
+      dtype: image
+  # --- Add config information if you have multiple (coco_far_voc, etc.) ---
+  # This helps Croissant understand different parts of your dataset
+  configs:
+    - config_name: coco_far_voc # Match your restructured folder names
+      # data_files: # Optional but can help Croissant
+      #   - split: train # Or whatever your split name is
+      #     path: coco_far_voc/metadata.jsonl # Path relative to dataset root
+      #   - split: train
+      #     path: coco_far_voc/* # For images
+    - config_name: coco_farther_bdd
+      # ... and so on for all 6 configurations
+    - config_name: oi_far_voc
+    - config_name: oi_farther_bdd
+    - config_name: oi_near_voc
+    - config_name: coco_near_voc
+
+---
+# Rest of your README.md content