[WIP] update files

MSCOCO.py

@@ -338,6 +338,7 @@ def compress_rle(
         height: int,
         width: int,
     ) -> CompressedRLE:
+        breakpoint()
         if iscrowd:
             rle = cocomask.frPyObjects(segmentation, h=height, w=width)
         else:
@@ -498,6 +499,52 @@ def from_dict(
         )
 
 
+@dataclass
+class StuffAnnotationData(InstancesAnnotationData):
+    @classmethod
+    def from_dict(
+        cls,
+        json_dict: JsonDict,
+        images: Dict[ImageId, ImageData],
+        decode_rle: bool,
+    ) -> "StuffAnnotationData":
+        segmentation = json_dict["segmentation"]
+        image_id = json_dict["image_id"]
+        image_data = images[image_id]
+        iscrowd = bool(json_dict["iscrowd"])
+
+        segmentation_mask = (
+            cls.rle_segmentation_to_mask(
+                segmentation=segmentation,
+                iscrowd=iscrowd,
+                height=image_data.height,
+                width=image_data.width,
+            )
+            if decode_rle
+            else cls.compress_rle(
+                segmentation=segmentation,
+                iscrowd=iscrowd,
+                height=image_data.height,
+                width=image_data.width,
+            )
+        )
+        return cls(
+            #
+            # for AnnotationData
+            #
+            annotation_id=json_dict["id"],
+            image_id=image_id,
+            #
+            # for InstancesAnnotationData
+            #
+            segmentation=segmentation_mask,
+            area=json_dict["area"],
+            iscrowd=iscrowd,
+            bbox=json_dict["bbox"],
+            category_id=json_dict["category_id"],
+        )
+
+
 class LicenseDict(TypedDict):
     license_id: LicenseId
     name: str
@@ -926,10 +973,21 @@ def split_generators(self, file_paths: Dict[str, Any]) -> List[ds.SplitGenerator
         ]
 
     def load_data(
-        self, ann_dicts: List[JsonDict], tqdm_desc: str = "Load stuff data", **kwargs
+        self,
+        ann_dicts: List[JsonDict],
+        images: Dict[ImageId, ImageData],
+        decode_rle: bool,
+        tqdm_desc: str = "Load stuff data",
     ):
         annotations = defaultdict(list)
-        breakpoint()
+        ann_dicts = sorted(ann_dicts, key=lambda d: d["image_id"])
+
+        for ann_dict in tqdm(ann_dicts, desc=tqdm_desc):
+            ann_data = StuffAnnotationData.from_dict(
+                ann_dict, images=images, decode_rle=decode_rle
+            )
+            annotations[ann_data.image_id].append(ann_data)
+        return annotations
 
     def generate_examples(
         self,