Tables in business documents power analytics and compliance, yet task-specific datasets are costly to build. Practitioners therefore turn to zero-shot vision–language models (VLMs). We study zero-shot realism for table detection (TD) and table structure recognition (TSR) under a unified protocol on DocILE-QUEST and a private STM154 corpus. We report TD with GIoU, Purity, and Completeness, and TSR with TEDS and TEDS-S, evaluating commercial VLMs (GPT-4o, GPT-5-mini), compact detectors, and supervised YOLO/DETR baselines. Zero-shot VLMs are strong for TSR and competitive for TD, while fine-tuned or from-scratch detectors lead when box quality and robustness to clutter matter. We add an automated error taxonomy that isolates actionable failures (missed, merged/split tables, header–body confusions, cell topology). Finally, we quantify emissions, finding a 104gap between the lightest and heaviest systems.

Document classification systems become more and more complex with the need to deal with new document formats or categories while obtaining a low error rate when classifying more and more documents. Such systems need to have important features including (1) the ability to eliminate ambiguity to improve precision or reduce the error rate, (2) the capability to detect and reject documents belonging to new categories or new variations. Previous studies often focused on closed datasets or solely on the problem of novelty detection, without evaluating the ability to reject ambiguous results after the novelty detection. In this paper, we propose an end-to-end document classification algorithm including both novelty and ambiguity rejection. The proposed algorithm utilizes deep metric learning to compact the knowledge space, and then uses the last hidden layer’s features as input for an unsupervised KNN-based method for novelty and ambiguity rejection. Extensive experiments and analysis on private and public benchmark datasets demonstrate the effectiveness of our proposed algorithm. The algorithm provides the capability to handle new documents while effectively rejecting ambiguity to enhance the precision, recall of known categories, and coverage rate of the end-to-end document classification system.

We propose both a new strategy to weight a text vector for document classification and a comparison of a deep learning approach versus an incremental classification approach, integrating our novel strategy. Bag-of-word vectors are classic approaches to describe a textual document in document classification objective. A weakness of the bag of words is to lose the organization of the discourse within the document. Inspired by some Deep Learning approaches and Natural Language Processing for text classification, we suggest a simple strategy, featuring the terms according to their relative positions within the discourse sequence. For experimentations, we apply this strategy to a recent document incremental classification approach from the state-of-the-art. And, we propose an original comparison between Incremental learning and Deep learning, by comparing the incremental system with a CCN-RNN-based approach. It demonstrates that both approaches are competitive in similar contest.

In this paper, we present an incremental frame-work for extracting information fields from administrative documents. First, we demonstrate some limits of the existing state-of-the-art methods such as the delay of the system efficiency. This is a concern in industrial context when we have only few samples of each document class. Based on this analysis, we propose a hybrid system combining incremental learning by means of itf-df statistics and a-priori generic models. We report in the experimental section our results obtained with a dataset of real invoices.

Our objective is to propose a Document Understanding System for Digital Mailroom application which can cope with three challenges: (1) process a full workflow with high accuracy, with the constraint of a partial training; (2) minimal requirement for configuration work from expert users; (3) adapt incrementally the system in quasi real-time to continuously maximize the recall. We describe an end-to-end system based on existing incremental algorithms for both document classification and field extraction. But in this paper, we really focus on the document classification issue. The main contribution is to adapt the Incremental Growing Neural Gas (A2ING) with a dynamic incremental feature vector. Moreover, a generic Framework automatically selects textual descriptors relying on performance. The quality assessment converges the A2ING and controls the system accuracy.

In this paper, we describe a method to restore a flow of continuous documents. The flow is a collection of consecutive scanned pages without explicit separation marks between documents. Our method is based on contextual and layout descriptors meant to specify the relationship between each pair of consecutive pages. The relationships are represented using vectors of features with boolean values indicating the presence or the absence of descriptors on concerned pages. The segmentation task therefore consists in classifying such vectors into continuities or breaks. The continuity class indicates that pages belong to the same document while the break class ends the ongoing document and starts a new one. The experimental part is based on a large collection of real administrative documents.

This paper describes the ICPR2018 fraud detection contest, its data set, its evaluation methodology, as well as the different methods submitted by the participants to tackle the predefined tasks. Forensics research is quite a sensitive topic. Data are either private or unlabeled and most of related works are evaluated on private datasets with a restricted access. This restriction has two major consequences: results cannot be reproduced and no benchmarking can be done between every approach. This contest was conceived in order to address these drawbacks. Two tasks were proposed: detecting documents containing at least one forgery in a flow of documents and spotting and localizing these forgeries within documents. An original dataset composed of images and texts of French receipts was provided to participants. The results they obtained are presented and discussed.

Document forgery is an increasing problem for both the public administration and private companies. It represents substantial losses in time and economical resources. Classical solutions to this problem such as watermarks or other integrated security patterns can not be applied in general for any unknown incoming document due to the large variability on types of documents. In that scenario it is important to resort to forensic techniques to seek and analyze inconsistencies on the intrinsic features of the document image. In this paper we present a classification-based approach for forgery detection. We use uniform Local Binary Patterns (LBP) to capture discriminant texture features that are common on forged regions. Besides, we combine multiple descriptors from neighboring regions to model contextual information. Results using Support Vector Machines (SVM) for patch classification show that we are able to detect several types of forgeries in a wide range of types of documents.

All indications show that paper documents will not cede in favour of their digital counterparts, but will instead be used increasingly in conjunction with digital information. An open challenge is how to seamlessly link the physical with the digital -- how to continue taking advantage of the important affordances of paper, without missing out on digital functionality. This paper presents the authors' experience with developing systems for Human-Document Interaction based on augmented document interfaces and examines new challenges and opportunities arising for the document image analysis field in this area. The system presented combines state of the art camera-based document image analysis techniques with a range of complementary technologies to offer fluid Human-Document Interaction. Both fixed and nomadic setups are discussed that have gone through user testing in real-life environments, and use cases are presented that span the spectrum from business to educational applications.

This paper proposes an entity local structure comparison approach based on inexact subgraph matching. The comparison results are used for mislabeling correction in the local structure. The latter represents a set of entity attribute labels which are physically close in a document image. It is modeled by an attributed graph describing the content and presentation features of the labels by the nodes and the geometrical features by the arcs. A local structure graph is matched with a structure model which represents a set of local structure model graphs. The structure model is initially built using a set of well chosen local structures based on a graph clustering algorithm and is then incrementally updated. The subgraph matching adopts a specific cost function that integrates the feature dissimilarities. The matched model graph is used to extract the missed labels, prune the extraneous ones and correct the erroneous label fields in the local structure. The evaluation of the structure comparison approach on 525 local structures extracted from 200 business documents achieves about 90% for recall and 95% for precision. The mislabeling correction rates in these local structures vary between 73% and 100%.

Document image classification in a professional context requires to respect some constraints such as dealing with a large variability of documents and/or number of classes. Whereas most methods deal with all classes at the same time, we answer this problem by presenting a new compliant system based on the specialization of the features and the parametrization of the classifier separately, class per class. We first compute a generalized vector of features based on global image characterization and structural primitives. Then, for each class, the feature vector is specialized by ranking the features according a stability score. Finally, a one-class K-nn classifier is trained using these specific features. Conducted experiments reveal good classification rates, proving the ability of our system to deal with a large range of documents classes.

Administrative document images are usually processed in black and white what generates many problems due to the errors related to the binarization. Besides all semantic information provided by the color is lost. Document images have a rich and highly variable content. The presence of false colors and artefacts introduced by the scanning and the compression alter the segmentation of the regions. Problems arise when there is no correspondence between the point clouds which are detected in a color space and the real regions of an image. In order to help the segmentation, we propose the extraction of the main colors of an image as a set of binary layers. Due to the industrial context, our approach has to run unsupervised on a generic dataset of color administrative documents. The originality of this approach is the use of a multiresolution analysis to detect the number of colors automatically. At a low resolution, a set of local regions is obtained thanks to a SLIC-based approach which takes into account the structure of documents and which combines both colorimetric information and spatial information. Then, a merging stage is applied on each resolution separately based on the colors which have been extracted at a lower resolution. This contribution can both feed the traditional process and exploit colorimetric information.

This paper presents an approach for spotting textual fields in commercial and administrative colored forms. We proceed by locating these fields thanks to their neighboring context which is modeled with a structural representation. First, informative zones are extracted. Second, forms are represented by graphs. In these graphs, nodes represent colored rectangular shapes while edges represent neighboring relations. Finally, the neighboring context of the queried region of interest is modeled as a graph. Subgraph isomorphism is applied in order to locate this ROI in the structural representation of a whole document. Evaluated on a 130-document image dataset, experimental results show up that our approach is efficient and that the requested information is found even if its position is changed.

This paper addresses the problem of machine printed and handwritten text separation in real noisy documents. We have proposed in a previous work a robust separation system relying on a proximity string segmentation algorithm. The extracted pseudo-lines and pseudo-words are used as basic blocks for classification. A multi-class support vector machine (SVM) with Gaussian kernel associates first an appropriate label to each pseudo-word. Then, the local neighborhood of each pseudo-word is studied in order to propagate the context and correct the classification errors. In this work, we first propose to model the separation problem by conditional random fields considering the horizontal neighborhood. As the considered neighborhood is too local to solve certain error cases, we have enhanced this method by using a more global context based on class dominance in the pseudo-line. The method has been evaluated on business documents. It separates handwritten and printed text with better scores (99.1% and 99.2% respectively), contrary to noise which is very random in these documents (90.1%).

We propose in this paper a framework for the segmentation and classification of document streams. The framework is composed of two modules: segmentation and verification. The two modules use an incremental classifier which learns progressively along the stream. In the segmentation module a relationship between two consecutive pages is classified as either: continuity or rupture. Rupture is synonymous of a clear break, thus probably a complete document. If the classifier is uncertain on whether the relationship should be a continuity or a rupture, an over-segmentation is proposed and we consider that we have a fragment i.e. portion of a document. Both fragments and documents are sent to the verification module where additionally to the incremental classifier it includes a correction module. The classifier predicts the classes of fragments and documents. The predicted class represents a context which is used as a query to search for similar contexts in the correction module and correct the segmentation and verification results. Corrections are sent back to the segmentation and verification modules to learn the correct classes. Results on real world databases show the effectiveness and stability of our approach.

In this paper we present an incremental framework aimed at extracting field information from administrative document images in the context of a Digital Mail-room scenario. Given a single training sample in which the user has marked which fields have to be extracted from a particular document class, a document model representing structural relationships among words is built. This model is incrementally refined as the system processes more and more documents from the same class. A reformulation of the tf-idf statistic scheme allows to adjust the importance weights of the structural relationships among words. We report in the experimental section our results obtained with a large dataset of real invoices.

Many symbols in music scores are linear segments. In this context, we designed an extractor of segments. It is robust towards problems of quality within binary images (scale factor, curvature, bias and noises). It is based on Kalman filtering technique. By splitting music scores into layers of detectable symbols and by applying methodically to the defined layers both this extractor and simple rules of classification for the detected segments, we were able to recognize staves, stems, slurs, beams, bar lines, black note heads and then quarters and note groups.