Sync DeepSeek OCR HF job code

ds_batch_ocr/document.py

@@ -19,6 +19,10 @@ GROUNDING_PATTERN = re.compile(
     re.DOTALL,
 )
 
+ALT_GROUNDING_PATTERN = re.compile(
+    r"(?P<label>[A-Za-z0-9_]+)\[\[(?P<coords>[^\[\]]+)\]\]",
+)
+
 FIGURE_MARKDOWN_PATTERN = re.compile(
     r"!\[Figure (?P<figure_id>[^\]]+)\]\((?P<path>[^)]+)\)"
 )
@@ -47,6 +51,25 @@ def extract_grounding_blocks(text: str) -> List[Dict[str, Any]]:
                 "span": match.span(),
             }
         )
+
+    if not matches:
+        for match in ALT_GROUNDING_PATTERN.finditer(text):
+            label = match.group("label").strip()
+            coords_text = match.group("coords").strip()
+            coordinates = None
+            if coords_text:
+                try:
+                    coordinates = ast.literal_eval(f"[{coords_text}]")
+                except Exception:
+                    coordinates = None
+            matches.append(
+                {
+                    "label": label,
+                    "coordinates": coordinates,
+                    "raw": match.group(0),
+                    "span": match.span(),
+                }
+            )
     return matches
 
 
@@ -55,11 +78,16 @@ def flatten_boxes(coordinates: Any) -> List[List[float]]:
     if coordinates is None:
         return boxes
     if isinstance(coordinates, (list, tuple)):
-        for item in coordinates:
-            if isinstance(item, (list, tuple)) and len(item) == 4:
-                boxes.append([float(value) for value in item])
-            elif isinstance(item, dict):
-                boxes.extend(flatten_boxes(item.get("bbox")))
+        if len(coordinates) == 4 and all(
+            isinstance(value, (int, float)) for value in coordinates
+        ):
+            boxes.append([float(value) for value in coordinates])
+        else:
+            for item in coordinates:
+                if isinstance(item, (list, tuple)) and len(item) == 4:
+                    boxes.append([float(value) for value in item])
+                elif isinstance(item, dict):
+                    boxes.extend(flatten_boxes(item.get("bbox")))
     elif isinstance(coordinates, dict):
         boxes.extend(flatten_boxes(coordinates.get("bbox")))
     return boxes
@@ -82,10 +110,24 @@ def clamp(value: int, lower: int, upper: int) -> int:
 def normalized_to_pixels(box: List[float], width: int, height: int) -> Optional[List[int]]:
     if len(box) != 4 or width <= 0 or height <= 0:
         return None
-    x1 = clamp(int(round(box[0] / 999.0 * width)), 0, width)
-    y1 = clamp(int(round(box[1] / 999.0 * height)), 0, height)
-    x2 = clamp(int(round(box[2] / 999.0 * width)), 0, width)
-    y2 = clamp(int(round(box[3] / 999.0 * height)), 0, height)
+
+    max_value = max(box)
+    min_value = min(box)
+
+    if max_value <= 1.0:
+        scale_x = width
+        scale_y = height
+    elif max_value <= 999.0:
+        scale_x = width / 999.0
+        scale_y = height / 999.0
+    else:
+        scale_x = 1.0
+        scale_y = 1.0
+
+    x1 = clamp(int(round(box[0] * scale_x)), 0, width)
+    y1 = clamp(int(round(box[1] * scale_y)), 0, height)
+    x2 = clamp(int(round(box[2] * scale_x)), 0, width)
+    y2 = clamp(int(round(box[3] * scale_y)), 0, height)
     if x2 <= x1 or y2 <= y1:
         return None
     return [x1, y1, x2, y2]
@@ -127,6 +169,21 @@ def save_figure(
     if not merged_box:
         return None
     width, height = image.size
+    converted_boxes: List[List[int]] = []
+    normalized_boxes: List[List[float]] = []
+    for box in boxes:
+        pixels = normalized_to_pixels(box, width, height)
+        if pixels:
+            converted_boxes.append(pixels)
+            normalized_boxes.append(
+                [
+                    pixels[0] / float(width),
+                    pixels[1] / float(height),
+                    pixels[2] / float(width),
+                    pixels[3] / float(height),
+                ]
+            )
+
     pixel_box = normalized_to_pixels(merged_box, width, height)
     if not pixel_box:
         return None
@@ -142,7 +199,12 @@ def save_figure(
     full_path = figures_dir / figure_filename
     crop.save(full_path)
 
-    norm_box = [value / 999.0 for value in merged_box]
+    norm_box = [
+        pixel_box[0] / float(width),
+        pixel_box[1] / float(height),
+        pixel_box[2] / float(width),
+        pixel_box[3] / float(height),
+    ]
     bounding_box_norm = {
         "x1": norm_box[0],
         "y1": norm_box[1],
@@ -157,7 +219,7 @@ def save_figure(
         image_path=(Path(sample_id) / figure_relative_doc_path).as_posix(),
         document_relative_path=figure_relative_doc_path.as_posix(),
         bounding_box_norm=bounding_box_norm,
-        bounding_box_norm_list=[[float(v) for v in box] for box in boxes],
+        bounding_box_norm_list=normalized_boxes or [[float(v) for v in norm_box]],
         bounding_box_pixels=bounding_box_pixels,
     )