An open-source document parsing model with SOTA accuracy and speed, using an RSI loop during post-training.
#!/usr/bin/env python3
"""Send one or more images to the D-FINE LA HTTP API."""
import argparse
from contextlib import ExitStack
import json
from pathlib import Path
import requests
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("images", nargs="+", type=Path)
parser.add_argument("--url", default="http://127.0.0.1:8000/inference")
parser.add_argument("--confidence", type=float)
parser.add_argument(
"--class-confidences",
help=(
"JSON object keyed by numeric class ID; for example "
"'{\"4\":0.7,\"8\":0.3}' sets Page-footer and Table overrides. "
"Unspecified classes inherit --confidence."
),
)
args = parser.parse_args()
data = {}
if args.confidence is not None:
data["confidence"] = str(args.confidence)
if args.class_confidences is not None:
data["class_confidences"] = args.class_confidences
with ExitStack() as stack:
files = [
(
"images",
(path.name, stack.enter_context(path.open("rb")), "application/octet-stream"),
)
for path in args.images
]
response = requests.post(args.url, files=files, data=data, timeout=120)
response.raise_for_status()
print(json.dumps(response.json(), ensure_ascii=False))
if __name__ == "__main__":
main()