ICDAR2019 Competition on Scanned Receipt OCR and Information Extraction

Abstract

Scanned receipts OCR and key information extraction (SROIE) represent the processeses of recognizing text from scanned receipts and extracting key texts from them and save the extracted tests to structured documents. SROIE plays critical roles for many document analysis applications and holds great commercial potentials, but very little research works and advances have been published in this area. In recognition of the technical challenges, importance and huge commercial potentials of SROIE, we organized the ICDAR 2019 competition on SROIE. In this competition, we set up three tasks, namely, Scanned Receipt Text Localisation (Task 1), Scanned Receipt OCR (Task 2) and Key Information Extraction from Scanned Receipts (Task 3). A new dataset with 1000 whole scanned receipt images and annotations is created for the competition. The competition opened on 10th February, 2019 and closed on 5th May, 2019. There are 29, 24 and 18 valide submissions received for the three competition tasks, respectively. In this report we will presents the motivation, competition datasets, task definition, evaluation protocol, submission statistics, performance of submitted methods and results analysis. According to the wide interests gained through SROIE and the healthy number of submissions from academic, research institutes and industry over different countries, we believe the competition SROIE is successful. And it is interesting to observe many new ideas and approaches are proposed for the new competition task set on key information extraction. According to the performance of the submissions, we believe there is still a large gap on the expected information extraction performance. Task of key information extraction is still very challenging and can be set for many other important document analysis applications. It is hoped that this competition will help draw more attention from the community and promote research and development efforts on SROIE.

LayoutLMv2: Multi-modal Pre-training for Visually-rich Document Understanding

“…Six datasets are used as down-stream tasks. The FUNSD (Jaume et al, 2019), CORD (Park et al, 2019), SROIE (Huang et al, 2019) and Kleister-NDA (Graliński et al, 2020) datasets define entity extraction tasks that aim to extract the value of a set of pre-defined keys, which we formalize as a sequential labeling task. RVL-CDIP (Harley et al, 2015) is for document image classification.…”

Efficient Automated Processing of the Unstructured Documents Using Artificial Intelligence: A Systematic Literature Review and Future Directions

“…However, registration for such datasets is typically required. For example, ICDAR has such types of datasets, on which many researchers have proposed their work [59], [74].…”

“…Text extraction is the main stage in automating document image processing [24], [32], [50], [109]. The document images can be compressed or uncompressed, grayscale or color and the text in the images can be editable or non-editable [64], [59], [39].…”

“…A range of information extraction techniques are proposed for particular applications, containing metadata extraction from scientific journals [3], legal contract entity extraction [100], [84], receipt entity extraction [59], and clinical text extraction [93], [20]. It is quite challenging to design a general-purpose text information extraction system as there are a lot of variations in a document image.…”

“…In this approach, a window of appropriate size, say n x m, is selected to search the desired text over the target image. Few studies [70], [87], [117], [84] Text Detector (EAST) [61], [59], [53] Feature extraction is learning and deriving the feature set, that accurately and distinctly describes the shape of a given character [90], [30]. Feature selection algorithms select the best feature subset from the input feature set [43].…”