# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """RVL-CDIP (Ryerson Vision Lab Complex Document Information Processing) dataset""" import os import numpy as np from tqdm import tqdm import datasets from pathlib import Path _CITATION = """\ @inproceedings{harley2015icdar, title = {Evaluation of Deep Convolutional Nets for Document Image Classification and Retrieval}, author = {Adam W Harley and Alex Ufkes and Konstantinos G Derpanis}, booktitle = {International Conference on Document Analysis and Recognition ({ICDAR})}}, year = {2015} } """ _DESCRIPTION = """\ The RVL-CDIP (Ryerson Vision Lab Complex Document Information Processing) dataset consists of 400,000 grayscale images in 16 classes, with 25,000 images per class. There are 320,000 training images, 40,000 validation images, and 40,000 test images. """ _HOMEPAGE = "https://www.cs.cmu.edu/~aharley/rvl-cdip/" _LICENSE = "https://www.industrydocuments.ucsf.edu/help/copyright/" _URLS = { "rvl-cdip": "https://huggingface.co/datasets/rvl_cdip/resolve/main/data/rvl-cdip.tar.gz", } _METADATA_URLS = { "train": "https://huggingface.co/datasets/rvl_cdip/resolve/main/data/train.txt", "test": "https://huggingface.co/datasets/rvl_cdip/resolve/main/data/test.txt", "val": "https://huggingface.co/datasets/rvl_cdip/resolve/main/data/val.txt", } _CLASSES = [ "letter", "form", "email", "handwritten", "advertisement", "scientific report", "scientific publication", "specification", "file folder", "news article", "budget", "invoice", "presentation", "questionnaire", "resume", "memo", ] _IMAGES_DIR = "images/" # hardcoded to not get stuck in annoying IO and LFS problems in Hub _OCR_DIR = "/cw/liir_data/NoCsBack/jordy/BDPC" _OCR_DIR = _OCR_DIR if os.path.exists(_OCR_DIR) else "data/" # class OCRConfig(datasets.BuilderConfig): # """BuilderConfig for RedCaps.""" # def __init__(self, name, OCR_dir, **kwargs): # """BuilderConfig for RedCaps. # Args: # **kwargs: keyword arguments forwarded to super. # """ # assert "description" not in kwargs # super(OCRConfig, self).__init__(version=kwargs["version"], name=name, **kwargs) # self.OCR_dir = OCR_dir class RvlCdipEasyOcr(datasets.GeneratorBasedBuilder): """Ryerson Vision Lab Complex Document Information Processing dataset.""" VERSION = datasets.Version("1.0.0") # BUILDER_CONFIGS = [OCRConfig("default",version=VERSION)] DEFAULT_CONFIG_NAME = "default" def _info(self): return datasets.DatasetInfo( description=_DESCRIPTION, features=datasets.Features( { "id": datasets.Value("string"), "image": datasets.Image(), "label": datasets.ClassLabel(names=_CLASSES), "words": datasets.Sequence(datasets.Value("string")), "boxes": datasets.Sequence( datasets.Sequence(datasets.Value("int32")) ), } ), homepage=_HOMEPAGE, citation=_CITATION, license=_LICENSE, ) def _split_generators(self, dl_manager): if self.config.data_files: archive_path = self.config.data_files["binary"][0] else: archive_path = dl_manager.download( _URLS["rvl-cdip"] ) # only download images if need be labels_path = dl_manager.download(_METADATA_URLS) return [ datasets.SplitGenerator( name=datasets.Split.TRAIN, gen_kwargs={ "archive_iterator": dl_manager.iter_archive(archive_path), "labels_filepath": labels_path["train"], "split": "train", }, ), datasets.SplitGenerator( name=datasets.Split.TEST, gen_kwargs={ "archive_iterator": dl_manager.iter_archive(archive_path), "labels_filepath": labels_path["test"], "split": "test", }, ), datasets.SplitGenerator( name=datasets.Split.VALIDATION, gen_kwargs={ "archive_iterator": dl_manager.iter_archive(archive_path), "labels_filepath": labels_path["val"], "split": "validation", }, ), ] @staticmethod def _get_image_to_class_map(data): image_to_class_id = {} for item in data: image_path, class_id = item.split(" ") image_path = os.path.join(_IMAGES_DIR, image_path) image_to_class_id[image_path] = int(class_id) return image_to_class_id @staticmethod def _get_image_to_OCR(OCR_dir, split): def parse_easyOCR_box(box): # {'x0': 39, 'y0': 39, 'x1': 498, 'y1': 82, 'width': 459, 'height': 43} return (box["x0"], box["y0"], box["x1"], box["y1"]) if OCR_dir is None: return {} image_to_OCR = {} data = np.load( os.path.join(OCR_dir, f"Easy_{split[0].upper()+split[1:]}_Data.npy"), allow_pickle=True, ) for ex in tqdm(data, desc="Loading OCR data"): w, h = ex["images"][0]["image_width"], ex["images"][0]["image_height"] filename = Path(ex["images"][0]["file_name"]).stem words = ex["word-level annotations"][0]["ocred_text"] box_info = ex["word-level annotations"][0]["ocred_boxes"] boxes = [parse_easyOCR_box(box) for box in box_info] assert len(boxes) == len(words) image_to_OCR[filename] = (words, boxes) return image_to_OCR @staticmethod def _path_to_OCR(image_to_OCR, file_path): # obtain text and boxes given file_path words, boxes = None, None #imagesv/v/u/b/vub13c00/523466896+-6898.tif #523466896+-6898.jpg file_path = Path(file_path).stem if file_path in image_to_OCR: words, boxes = image_to_OCR[file_path] return words, boxes def _generate_examples(self, archive_iterator, labels_filepath, split): with open(labels_filepath, encoding="utf-8") as f: data = f.read().splitlines() image_to_OCR = self._get_image_to_OCR(_OCR_DIR, split) image_to_class_id = self._get_image_to_class_map(data) for file_path, file_obj in archive_iterator: if file_path.startswith(_IMAGES_DIR): if file_path in image_to_class_id: class_id = image_to_class_id[file_path] label = _CLASSES[class_id] words, boxes = self._path_to_OCR(image_to_OCR, file_path) if words is not None: #skipping all items for which we do not have OCR a = dict( id=file_path, image={"path": file_path, "bytes": file_obj.read()}, label=label, words=words, boxes=boxes, ) yield file_path, a