def create_annotation_for_annotator(self,
anno_files: AnnotationFiles) -> None:
# Firstly initialize the annotation tables with the annotator name
annotator = anno_files.annotator
for token_data in self.all_page_token_data.values():
token_data[annotator] = None
pbar = tqdm(anno_files)
for anno_file in pbar:
paper_sha = anno_file["paper_sha"]
df = self.all_page_token_data[paper_sha]
pawls_annotations = load_json(
anno_file["annotation_path"])["annotations"]
for anno in pawls_annotations:
if anno["tokens"] is None:
continue
# Skip if current category is not in the specified categories
label = anno["label"]["text"]
if label not in self.categories:
continue
anno_token_indices = [(ele["pageIndex"], ele["tokenIndex"])
for ele in anno["tokens"]]
df.loc[anno_token_indices, annotator] = label
def load_directory(self, pred_dir: str) -> Dict[str, Dict]:
pdf_preds = {}
for pred_json in glob(f"{pred_dir}/*.json"):
filename = os.path.basename(pred_json).replace(".json", ".pdf")
self.pdf_preds[filename] = load_json(pred_json)
return pdf_preds
def create_annotation_for_annotator(self,
anno_files: AnnotationFiles) -> None:
"""Create the annotations for the given annotation files"""
_annotations = []
anno_id = 0
pbar = tqdm(anno_files)
for anno_file in pbar:
paper_sha = anno_file["paper_sha"]
pbar.set_description(f"Working on {paper_sha[:10]}...")
pawls_annotations = load_json(
anno_file["annotation_path"])["annotations"]
for anno in pawls_annotations:
page_id = anno["page"]
category = anno["label"]["text"]
# Skip if current category is not in the specified categories
cat_id = self._name2catid.get(category, None)
if cat_id is None:
continue
image_data = self.get_image_data(paper_sha, page_id)
width, height = image_data["width"], image_data["height"]
x, y, w, h = _convert_bounds_to_coco_bbox(anno["bounds"])
_annotations.append(
self.AnnoTemplate(
id=anno_id,
bbox=[x, y, w, h],
category_id=cat_id,
image_id=image_data["id"],
area=w * h,
)._asdict())
anno_id += 1
return _annotations
def __init__(self, pred_file: str):
if os.path.isfile(pred_file):
self.pdf_preds = load_json(pred_file)
elif os.path.isdir(pred_file):
self.pdf_preds = self.load_directory(pred_file)
