Spaces:
Running
Running
File size: 4,733 Bytes
f3270e6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 |
# Copyright (C) 2021-2025, Mindee.
# This program is licensed under the Apache License 2.0.
# See LICENSE or go to <https://opensource.org/licenses/Apache-2.0> for full license details.
import json
import os
from pathlib import Path
from typing import Any
import numpy as np
from tqdm import tqdm
from .datasets import VisionDataset
from .utils import convert_target_to_relative, crop_bboxes_from_image
__all__ = ["FUNSD"]
class FUNSD(VisionDataset):
"""FUNSD dataset from `"FUNSD: A Dataset for Form Understanding in Noisy Scanned Documents"
<https://arxiv.org/pdf/1905.13538.pdf>`_.
.. image:: https://doctr-static.mindee.com/models?id=v0.5.0/funsd-grid.png&src=0
:align: center
>>> from doctr.datasets import FUNSD
>>> train_set = FUNSD(train=True, download=True)
>>> img, target = train_set[0]
Args:
train: whether the subset should be the training one
use_polygons: whether polygons should be considered as rotated bounding box (instead of straight ones)
recognition_task: whether the dataset should be used for recognition task
detection_task: whether the dataset should be used for detection task
**kwargs: keyword arguments from `VisionDataset`.
"""
URL = "https://guillaumejaume.github.io/FUNSD/dataset.zip"
SHA256 = "c31735649e4f441bcbb4fd0f379574f7520b42286e80b01d80b445649d54761f"
FILE_NAME = "funsd.zip"
def __init__(
self,
train: bool = True,
use_polygons: bool = False,
recognition_task: bool = False,
detection_task: bool = False,
**kwargs: Any,
) -> None:
super().__init__(
self.URL,
self.FILE_NAME,
self.SHA256,
True,
pre_transforms=convert_target_to_relative if not recognition_task else None,
**kwargs,
)
if recognition_task and detection_task:
raise ValueError(
"`recognition_task` and `detection_task` cannot be set to True simultaneously. "
+ "To get the whole dataset with boxes and labels leave both parameters to False."
)
self.train = train
np_dtype = np.float32
# Use the subset
subfolder = os.path.join("dataset", "training_data" if train else "testing_data")
# # list images
tmp_root = os.path.join(self.root, subfolder, "images")
self.data: list[tuple[str | np.ndarray, str | dict[str, Any] | np.ndarray]] = []
for img_path in tqdm(
iterable=os.listdir(tmp_root), desc="Preparing and Loading FUNSD", total=len(os.listdir(tmp_root))
):
# File existence check
if not os.path.exists(os.path.join(tmp_root, img_path)):
raise FileNotFoundError(f"unable to locate {os.path.join(tmp_root, img_path)}")
stem = Path(img_path).stem
with open(os.path.join(self.root, subfolder, "annotations", f"{stem}.json"), "rb") as f:
data = json.load(f)
_targets = [
(word["text"], word["box"])
for block in data["form"]
for word in block["words"]
if len(word["text"]) > 0
]
text_targets, box_targets = zip(*_targets)
if use_polygons:
# xmin, ymin, xmax, ymax -> (x, y) coordinates of top left, top right, bottom right, bottom left corners
box_targets = [ # type: ignore[assignment]
[
[box[0], box[1]],
[box[2], box[1]],
[box[2], box[3]],
[box[0], box[3]],
]
for box in box_targets
]
if recognition_task:
crops = crop_bboxes_from_image(
img_path=os.path.join(tmp_root, img_path), geoms=np.asarray(box_targets, dtype=np_dtype)
)
for crop, label in zip(crops, list(text_targets)):
# filter labels with unknown characters
if not any(char in label for char in ["☑", "☐", "\u03bf", "\uf703", "\uf702", " "]):
self.data.append((crop, label.replace("–", "-")))
elif detection_task:
self.data.append((img_path, np.asarray(box_targets, dtype=np_dtype)))
else:
self.data.append((
img_path,
dict(boxes=np.asarray(box_targets, dtype=np_dtype), labels=list(text_targets)),
))
self.root = tmp_root
def extra_repr(self) -> str:
return f"train={self.train}"
|