• Encoder-only Transformer baseline for handwriting text recognition (HTR). • Added writer-style head tracks & controls user-identifiable memorization. • Neural-activation pruning removes writer clues while preserving HTR accuracy. • Writer-ID Confusion (WIC) enforces uniform IDs, boosting efficient machine unlearning. • Membership-inference tests confirm privacy gains without costly retraining. Handwritten Text Recognition (HTR) is crucial for document digitization, but handwritten data can contain user-identifiable features, like unique writing styles, posing privacy risks. Regulations such as the “right to be forgotten” require models to remove these sensitive traces without full retraining. We introduce a practical encoder-only transformer baseline as a robust reference for future HTR research. Building on this, we propose a two-stage unlearning framework for multihead transformer HTR models. Our method combines neural pruning with machine unlearning applied to a writer classification head, ensuring sensitive information is removed while preserving the recognition head. We also present Writer-ID Confusion (WIC), a method that forces the forget set to follow a uniform distribution over writer identities, unlearning user-specific cues while maintaining text recognition performance. We compare WIC to Random Labeling, Fisher Forgetting, Amnesiac Unlearning, and DELETE within our prune-unlearn pipeline and consistently achieve better privacy and accuracy trade-offs. This is the first systematic study of machine unlearning for HTR. Using metrics such as Accuracy, Character Error Rate (CER), Word Error Rate (WER), and Membership Inference Attacks (MIA) on the IAM and CVL datasets, we demonstrate that our method achieves state-of-the-art or superior performance for effective unlearning. These experiments show that our approach effectively safeguards privacy without compromising accuracy, opening new directions for document analysis research. Our code is publicly available at https://github.com/leitro/WIC-WriterIDConfusion-MachineUnlearning .

Document Visual Question Answering (DocVQA) has introduced a new paradigm for end-to-end document understanding, and quickly became one of the standard benchmarks for multimodal LLMs. Automating document processing workflows, driven by DocVQA models, presents significant potential for many business sectors. However, documents tend to contain highly sensitive information, raising concerns about privacy risks associated with training such DocVQA models. One significant privacy vulnerability, exploited by the membership inference attack, is the possibility for an adversary to determine if a particular record was part of the model's training data. In this paper, we introduce two novel membership inference attacks tailored specifically to DocVQA models. These attacks are designed for two different adversarial scenarios: a white-box setting, where the attacker has full access to the model architecture and parameters, and a black-box setting, where only the model's outputs are available. Notably, our attacks assume the adversary lacks access to auxiliary datasets, which is more realistic in practice but also more challenging. Our unsupervised methods outperform existing state-of-the-art membership inference attacks across a variety of DocVQA models and datasets, demonstrating their effectiveness and highlighting the privacy risks in this domain.

We propose DocVXQA, a novel framework for visually self-explainable document question answering, where the goal is not only to produce accurate answers to questions but also to learn visual heatmaps that highlight critical regions, offering interpretable justifications for the model decision. To integrate explanations into the learning process, we quantitatively formulate explainability principles as explicit learning criteria. Unlike conventional relevance map methods that solely emphasize regions relevant to the answer, our context-aware DocVXQA delivers explanations that are contextually sufficient yet representation-efficient. This fosters user trust while achieving a balance between predictive performance and interpretability in document visual question answering applications. Extensive experiments, including human evaluation, provide strong evidence supporting the effectiveness of our method.

Multi‑page Document Visual Question Answering (MP‑DocVQA) remains challenging because long documents not only strain computational resources but also reduce the effectiveness of the attention mechanism in large vision–language models (LVLMs). We tackle these issues with an Adaptive Visual In‑document Retrieval (AVIR) framework. A lightweight retrieval model first scores each page for question relevance. Pages are then clustered according to the score distribution to adaptively select relevant content. The clustered pages are screened again by Top-K to keep the context compact. However, for short documents, clustering reliability decreases, so we use a relevance probability threshold to select pages. The selected pages alone are fed to a frozen LVLM for answer generation, eliminating the need for model fine‑tuning. The proposed AVIR framework reduces the average page count required for question answering by 70%, while achieving an ANLS of 84.58% on the MP-DocVQA dataset—surpassing previous methods with significantly lower computational cost. The effectiveness of the proposed AVIR is also verified on the SlideVQA and DUDE benchmarks. Our code will be made publicly available upon acceptance.

Fonts are integral to creative endeavors, design processes, and artistic productions. The appropriate selection of a font can significantly enhance artwork and endow advertisements with a higher level of expressivity. Despite the availability of numerous diverse font designs online, traditional retrieval-based methods for font selection are increasingly being supplanted by generation-based approaches. These newer methods offer enhanced flexibility, catering to specific user preferences and capturing unique stylistic impressions. However, current impression font techniques based on Generative Adversarial Networks (GANs) necessitate the utilization of multiple auxiliary losses to provide guidance during generation. Furthermore, these methods commonly employ weighted summation for the fusion of impression-related keywords. This leads to generic vectors with the addition of more impression keywords, ultimately lacking in detail generation capacity. In this paper, we introduce a diffusion-based method, termed GRIF-DM, to generate fonts that vividly embody specific impressions, utilizing an input consisting of a single letter and a set of descriptive impression keywords. The core innovation of GRIF-DM lies in the development of dual cross-attention modules, which process the characteristics of the letters and impression keywords independently but synergistically, ensuring effective integration of both types of information. Our experimental results, conducted on the MyFonts dataset, affirm that this method is capable of producing realistic, vibrant, and high-fidelity fonts that are closely aligned with user specifications. This confirms the potential of our approach to revolutionize font generation by accommodating a broad spectrum of user-driven design requirements. Our code is publicly available at https://github.com/leitro/GRIF-DM.

The comic domain is rapidly advancing with the development of single-page analysis and synthesis models. However, evaluation metrics and datasets lag behind, often limited to small-scale or single-style test sets. We introduce a novel benchmark, CoMix, designed to evaluate the multi-task capabilities of models in comic analysis. Unlike existing benchmarks that focus on isolated tasks such as object detection or text recognition, CoMix addresses a broader range of tasks including object detection, speaker identification, character re-identification, reading order, and multi-modal reasoning tasks like character naming and dialogue generation. Our benchmark comprises three existing datasets with expanded annotations to support multi-task evaluation. To mitigate the over-representation of manga-style data, we have incorporated a new dataset of carefully selected American comic-style books, thereby enriching the diversity of comic styles. CoMix is designed to assess pre-trained models in zero-shot and limited fine-tuning settings, probing their transfer capabilities across different comic styles and tasks. The validation split of the benchmark is publicly available for research purposes, and an evaluation server for the held-out test split is also provided. Comparative results between human performance and state-of-the-art models reveal a significant performance gap, highlighting substantial opportunities for advancements in comic understanding. The dataset, baseline models, and code are accessible at https://github.com/emanuelevivoli/CoMix-dataset. This initiative sets a new standard for comprehensive comic analysis, providing the community with a common benchmark for evaluation on a large and varied set.

We introduce the structured scene-text spotting task, which requires a scene-text OCR system to spot text in the wild according to a query regular expression. Contrary to generic scene-text OCR, structured scene-text spotting seeks to dynamically condition both detection and recognition on user-provided regular expressions. To tackle this task, we propose the Structured TExt sPotter (STEP), a model that exploits the provided text structure to guide the OCR process. STEP is able to deal with regular expressions that contain spaces and it is not bound to detection at wordlevel granularity. Our approach enables accurate zero-shot structured text spotting in a wide variety of real-world reading scenarios and is solely trained on publicly available data. To demonstrate the effectiveness of our approach, we introduce a new challenging test dataset that contains several types of out-of-vocabulary structured text, reflecting important reading applications such as weight information, serial numbers, license plates etc. We demonstrate that STEP can provide specialized OCR performance on demand in all tested scenarios. The code and test dataset are released at https://github.com/CVC-DAG/STEP.

Humans exploit prior knowledge to describe images, and are able to adapt their explanation to specific contextual information given, even to the extent of inventing plausible explanations when contextual information and images do not match. In this work, we propose the novel task of captioning Wikipedia images by integrating contextual knowledge. Specifically, we produce models that jointly reason over Wikipedia articles, Wikimedia images and their associated descriptions to produce contextualized captions. The same Wikimedia image can be used to illustrate different articles, and the produced caption needs to be adapted to the specific context allowing us to explore the limits of the model to adjust captions to different contextual information. Dealing with out-of-dictionary words and Named Entities is a challenging task in this domain. To address this, we propose a pre-training objective, Masked Named Entity Modeling (MNEM), and show that this pretext task results to significantly improved models. Furthermore, we verify that a model pre-trained in Wikipedia generalizes well to News Captioning datasets. We further define two different test splits according to the difficulty of the captioning task. We offer insights on the role and the importance of each modality and highlight the limitations of our model.

In this paper, we propose a Text-Degradation Invariant Auto Encoder (Text-DIAE), a self-supervised model designed to tackle two tasks, text recognition (handwritten or scene-text) and document image enhancement. We start by employing a transformer-based architecture that incorporates three pretext tasks as learning objectives to be optimized during pre-training without the usage of labelled data. Each of the pretext objectives is specifically tailored for the final downstream tasks. We conduct several ablation experiments that confirm the design choice of the selected pretext tasks. Importantly, the proposed model does not exhibit limitations of previous state-of-the-art methods based on contrastive losses, while at the same time requiring substantially fewer data samples to converge. Finally, we demonstrate that our method surpasses the state-of-the-art in existing supervised and self-supervised settings in handwritten and scene text recognition and document image enhancement. Our code and trained models will be made publicly available at https://github.com/dali92002/SSL-OCR

Video Question Answering methods focus on common-sense reasoning and visual cognition of objects or persons and their interactions over time. Current VideoQA approaches ignore the textual information present in the video. Instead, we argue that textual information is complementary to the action and provides essential contextualisation cues to the reasoning process. To this end, we propose a novel VideoQA task that requires reading and understanding the text in the video. To explore this direction, we focus on news videos and require QA systems to comprehend and answer questions about the topics presented by combining visual and textual cues in the video. We introduce the "NewsVideoQA" dataset that comprises more than 8, 600 QA pairs on 3, 000+ news videos obtained from diverse news channels from around the world. We demonstrate the limitations of current Scene Text VQA and VideoQA methods and propose ways to incorporate scene text information into VideoQA methods.

Explaining an image with missing or non-existent objects is known as object bias (hallucination) in image captioning. This behaviour is quite common in the state-of-the-art captioning models which is not desirable by humans. To decrease the object hallucination in captioning, we propose three simple yet efficient training augmentation method for sentences which requires no new training data or increase in the model size. By extensive analysis, we show that the proposed methods can significantly diminish our models’ object bias on hallucination metrics. Moreover, we experimentally demonstrate that our methods decrease the dependency on the visual features. All of our code, configuration files and model weights are available online1.

The task of image-text matching aims to map representations from different modalities into a common joint visual-textual embedding. However, the most widely used datasets for this task, MSCOCO and Flickr30K, are actually image captioning datasets that offer a very limited set of relation-ships between images and sentences in their ground-truth annotations. This limited ground truth information forces us to use evaluation metrics based on binary relevance: given a sentence query we consider only one image as relevant. However, many other relevant images or captions may be present in the dataset. In this work, we propose two metrics that evaluate the degree of semantic relevance of retrieved items, independently of their annotated binary relevance. Additionally, we incorporate a novel strategy that uses an image captioning metric, CIDEr, to define a Semantic Adaptive Margin (SAM) to be optimized in a standard triplet loss. By incorporating our formulation to existing models, a large improvement is obtained in scenarios where available training data is limited. We also demonstrate that the performance on the annotated image-caption pairs is maintained while improving on other non-annotated relevant items when employing the full training set. The code for our new metric can be found at github.com/furkanbiten/ncs_metric and the model implementation at github.com/andrespmd/semantic_adaptive_margin.

Infographics communicate information using a combination of textual, graphical and visual elements. This work explores the automatic understanding of infographic images by using a Visual Question Answering technique. To this end, we present InfographicVQA, a new dataset comprising a diverse collection of infographics and question-answer annotations. The questions require methods that jointly reason over the document layout, textual content, graphical elements, and data visualizations. We curate the dataset with an emphasis on questions that require elementary reasoning and basic arithmetic skills. For VQA on the dataset, we evaluate two Transformer-based strong baselines. Both the baselines yield unsatisfactory results compared to near perfect human performance on the dataset. The results suggest that VQA on infographics—images that are designed to communicate information quickly and clearly to human brain—is ideal for benchmarking machine understanding of complex document images. The dataset is available for download at docvqa.org

Low resource Handwritten Text Recognition (HTR) is a hard problem due to the scarce annotated data and the very limited linguistic information (dictionaries and language models). For example, in the case of historical ciphered manuscripts, which are usually written with invented alphabets to hide the message contents. Thus, in this paper we address this problem through a data generation technique based on Bayesian Program Learning (BPL). Contrary to traditional generation approaches, which require a huge amount of annotated images, our method is able to generate human-like handwriting using only one sample of each symbol in the alphabet. After generating symbols, we create synthetic lines to train state-of-the-art HTR architectures in a segmentation free fashion. Quantitative and qualitative analyses were carried out and confirm the effectiveness of the proposed method.

Scene text instances found in natural images carry explicit semantic information that can provide important cues to solve a wide array of computer vision problems. In this paper, we focus on leveraging multi-modal content in the form of visual and textual cues to tackle the task of fine-grained image classification and retrieval. First, we obtain the text instances from images by employing a text reading system. Then, we combine textual features with salient image regions to exploit the complementary information carried by the two sources. Specifically, we employ a Graph Convolutional Network to perform multi-modal reasoning and obtain relationship-enhanced features by learning a common semantic space between salient objects and text found in an image. By obtaining an enhanced set of visual and textual features, the proposed model greatly outperforms previous state-of-the-art in two different tasks, fine-grained classification and image retrieval in the Con-Text[23] and Drink Bottle[4] datasets.

Recent models for cross-modal retrieval have benefited from an increasingly rich understanding of visual scenes, afforded by scene graphs and object interactions to mention a few. This has resulted in an improved matching between the visual representation of an image and the textual representation of its caption. Yet, current visual representations overlook a key aspect: the text appearing in images, which may contain crucial information for retrieval. In this paper, we first propose a new dataset that allows exploration of cross-modal retrieval where images contain scene-text instances. Then, armed with this dataset, we describe several approaches which leverage scene text, including a better scene-text aware cross-modal retrieval method which uses specialized representations for text from the captions and text from the visual scene, and reconcile them in a common embedding space. Extensive experiments confirm that cross-modal retrieval approaches benefit from scene text and highlight interesting research questions worth exploring further. Dataset and code are available at europe.naverlabs.com/stacmr.

We present a new dataset for Visual Question Answering (VQA) on document images called DocVQA. The dataset consists of 50,000 questions defined on 12,000+ document images. Detailed analysis of the dataset in comparison with similar datasets for VQA and reading comprehension is presented. We report several baseline results by adopting existing VQA and reading comprehension models. Although the existing models perform reasonably well on certain types of questions, there is large performance gap compared to human performance (94.36% accuracy). The models need to improve specifically on questions where understanding structure of the document is crucial. The dataset, code and leaderboard are available at docvqa.org.

We present a method for exploiting weakly annotated images to improve text extraction pipelines. The approach uses an arbitrary end-to-end text recognition system to obtain text region proposals and their, possibly erroneous, transcriptions. The method includes matching of imprecise transcriptions to weak annotations and an edit distance guided neighbourhood search. It produces nearly error-free, localised instances of scene text, which we treat as “pseudo ground truth” (PGT). The method is applied to two weakly-annotated datasets. Training with the extracted PGT consistently improves the accuracy of a state of the art recognition model, by 3.7% on average, across different benchmark datasets (image domains) and 24.5% on one of the weakly annotated datasets11Acknowledgements. The authors were supported by Czech Technical University student grant SGS20/171/0HK3/3TJ13, the MEYS VVV project CZ.02.1.01/0.010.0J16 019/0000765 Research Center for Informatics, the Spanish Research project TIN2017-89779-P and the CERCA Programme / Generalitat de Catalunya.

Perceiving text is crucial to understand semantics of outdoor scenes and hence is a critical requirement to build intelligent systems for driver assistance and self-driving. Most of the existing datasets for text detection and recognition comprise still images and are mostly compiled keeping text in mind. This paper introduces a new "RoadText-1K" dataset for text in driving videos. The dataset is 20 times larger than the existing largest dataset for text in videos. Our dataset comprises 1000 video clips of driving without any bias towards text and with annotations for text bounding boxes and transcriptions in every frame. State of the art methods for text detection, recognition and tracking are evaluated on the new dataset and the results signify the challenges in unconstrained driving videos compared to existing datasets. This suggests that RoadText-1K is suited for research and development of reading systems, robust enough to be incorporated into more complex downstream tasks like driver assistance and self-driving. The dataset can be found at http://cvit.iiit.ac.in/research/projects/cvit-projects/roadtext-1k.

Image to image translation aims to learn a mapping that transforms an image from one visual domain to another. Recent works assume that images descriptors can be disentangled into a domain-invariant content representation and a domain-specific style representation. Thus, translation models seek to preserve the content of source images while changing the style to a target visual domain. However, synthesizing new images is extremely challenging especially in multi-domain translations, as the network has to compose content and style to generate reliable and diverse images in multiple domains. In this paper we propose the use of an image retrieval system to assist the image-to-image translation task. First, we train an image-to-image translation model to map images to multiple domains. Then, we train an image retrieval model using real and generated images to find images similar to a query one in content but in a different domain. Finally, we exploit the image retrieval system to fine-tune the image-to-image translation model and generate higher quality images. Our experiments show the effectiveness of the proposed solution and highlight the contribution of the retrieval network, which can benefit from additional unlabeled data and help image-to-image translation models in the presence of scarce data.

In this work we target the problem of hate speech detection in multimodal publications formed by a text and an image. We gather and annotate a large scale dataset from Twitter, MMHS150K, and propose different models that jointly analyze textual and visual information for hate speech detection, comparing them with unimodal detection. We provide quantitative and qualitative results and analyze the challenges of the proposed task. We find that, even though images are useful for the hate speech detection task, current multimodal models cannot outperform models analyzing only text. We discuss why and open the field and the dataset for further research.

Text contained in an image carries high-level semantics that can be exploited to achieve richer image understanding. In particular, the mere presence of text provides strong guiding content that should be employed to tackle a diversity of computer vision tasks such as image retrieval, fine-grained classification, and visual question answering. In this paper, we address the problem of fine-grained classification and image retrieval by leveraging textual information along with visual cues to comprehend the existing intrinsic relation between the two modalities. The novelty of the proposed model consists of the usage of a PHOC descriptor to construct a bag of textual words along with a Fisher Vector Encoding that captures the morphology of text. This approach provides a stronger multimodal representation for this task and as our experiments demonstrate, it achieves state-of-the-art results on two different tasks, fine-grained classification and image retrieval. The code of this model will be publicly available at1.

Current image captioning systems perform at a merely descriptive level, essentially enumerating the objects in the scene and their relations. Humans, on the contrary, interpret images by integrating several sources of prior knowledge of the world. In this work, we aim to take a step closer to producing captions that offer a plausible interpretation of the scene, by integrating such contextual information into the captioning pipeline. For this we focus on the captioning of images used to illustrate news articles. We propose a novel captioning method that is able to leverage contextual information provided by the text of news articles associated with an image. Our model is able to selectively draw information from the article guided by visual cues, and to dynamically extend the output dictionary to out-of-vocabulary named entities that appear in the context source. Furthermore we introduce ``GoodNews'', the largest news image captioning dataset in the literature and demonstrate state-of-the-art results.

Current visual question answering datasets do not consider the rich semantic information conveyed by text within an image. In this work, we present a new dataset, ST-VQA, that aims to highlight the importance of exploiting high-level semantic information present in images as textual cues in the Visual Question Answering process. We use this dataset to define a series of tasks of increasing difficulty for which reading the scene text in the context provided by the visual information is necessary to reason and generate an appropriate answer. We propose a new evaluation metric for these tasks to account both for reasoning errors as well as shortcomings of the text recognition module. In addition we put forward a series of baseline methods, which provide further insight to the newly released dataset, and set the scene for further research.

Word spotting has gained increased attention lately as it can be used to extract textual information from handwritten documents and scene-text images. Current word spotting approaches are designed to work on a single language and/or script. Building intelligent models that learn script-independent multilingual word-spotting is challenging due to the large variability of multilingual alphabets and symbols. We used ResNet-152 and the Pyramidal Histogram of Characters (PHOC) embedding to build a one-model script-independent multilingual word-spotting and we tested it on Latin, Arabic, and Bangla (Indian) languages. The one-model we propose performs on par with the multi-model language-specific word-spotting system, and thus, reduces the number of models needed for each script and/or language.

This paper presents final results of ICDAR 2019 Scene Text Visual Question Answering competition (ST-VQA). ST-VQA introduces an important aspect that is not addressed by any Visual Question Answering system up to date, namely the incorporation of scene text to answer questions asked about an image. The competition introduces a new dataset comprising 23,038 images annotated with 31,791 question / answer pairs where the answer is always grounded on text instances present in the image. The images are taken from 7 different public computer vision datasets, covering a wide range of scenarios. The competition was structured in three tasks of increasing difficulty, that require reading the text in a scene and understanding it in the context of the scene, to correctly answer a given question. A novel evaluation metric is presented, which elegantly assesses both key capabilities expected from an optimal model: text recognition and image understanding. A detailed analysis of results from different participants is showcased, which provides insight into the current capabilities of VQA systems that can read. We firmly believe the dataset proposed in this challenge will be an important milestone to consider towards a path of more robust and general models that can exploit scene text to achieve holistic image understanding.

This paper reports the ICDAR2019 Robust Reading Challenge on Arbitrary-Shaped Text - RRC-ArT that consists of three major challenges: i) scene text detection, ii) scene text recognition, and iii) scene text spotting. A total of 78 submissions from 46 unique teams/individuals were received for this competition. The top performing score of each challenge is as follows: i) T1 - 82.65%, ii) T2.1 - 74.3%, iii) T2.2 - 85.32%, iv) T3.1 - 53.86%, and v) T3.2 - 54.91%. Apart from the results, this paper also details the ArT dataset, tasks description, evaluation metrics and participants' methods. The dataset, the evaluation kit as well as the results are publicly available at the challenge website.

This paper explores the possibilities of image style transfer applied to text maintaining the original transcriptions. Results on different text domains (scene text, machine printed text and handwritten text) and cross-modal results demonstrate that this is feasible, and open different research lines. Furthermore, two architectures for selective style transfer, which means transferring style to only desired image pixels, are proposed. Finally, scene text selective style transfer is evaluated as a data augmentation technique to expand scene text detection datasets, resulting in a boost of text detectors performance. Our implementation of the described models is publicly available.

The ICDAR 2019 Challenge on "Scanned receipts OCR and key information extraction" (SROIE) covers important aspects related to the automated analysis of scanned receipts. The SROIE tasks play a key role in many document analysis systems and hold significant commercial potential. Although a lot of work has been published over the years on administrative document analysis, the community has advanced relatively slowly, as most datasets have been kept private. One of the key contributions of SROIE to the document analysis community is to offer a first, standardized dataset of 1000 whole scanned receipt images and annotations, as well as an evaluation procedure for such tasks. The Challenge is structured around three tasks, namely Scanned Receipt Text Localization (Task 1), Scanned Receipt OCR (Task 2) and Key Information Extraction from Scanned Receipts (Task 3). The competition opened on 10th February, 2019 and closed on 5th May, 2019. We received 29, 24 and 18 valid submissions received for the three competition tasks, respectively. This report presents the competition datasets, define the tasks and the evaluation protocols, offer detailed submission statistics, as well as an analysis of the submitted performance. While the tasks of text localization and recognition seem to be relatively easy to tackle, it is interesting to observe the variety of ideas and approaches proposed for the information extraction task. According to the submissions' performance we believe there is still margin for improving information extraction performance, although the current dataset would have to grow substantially in following editions. Given the success of the SROIE competition evidenced by the wide interest generated and the healthy number of submissions from academic, research institutes and industry over different countries, we consider that the SROIE competition can evolve into a useful resource for the community, drawing further attention and promoting research and development efforts in this field.

With the growing cosmopolitan culture of modern cities, the need of robust Multi-Lingual scene Text (MLT) detection and recognition systems has never been more immense. With the goal to systematically benchmark and push the state-of-the-art forward, the proposed competition builds on top of the RRC-MLT-2017 with an additional end-to-end task, an additional language in the real images dataset, a large scale multi-lingual synthetic dataset to assist the training, and a baseline End-to-End recognition method. The real dataset consists of 20,000 images containing text from 10 languages. The challenge has 4 tasks covering various aspects of multi-lingual scene text: (a) text detection, (b) cropped word script classification, (c) joint text detection and script classification and (d) end-to-end detection and recognition. In total, the competition received 60 submissions from the research and industrial communities. This paper presents the dataset, the tasks and the findings of the presented RRC-MLT-2019 challenge.

Robust text reading from street view images provides valuable information for various applications. Performance improvement of existing methods in such a challenging scenario heavily relies on the amount of fully annotated training data, which is costly and in-efficient to obtain. To scale up the amount of training data while keeping the labeling procedure cost-effective, this competition introduces a new challenge on Large-scale Street View Text with Partial Labeling (LSVT), providing 5,0000 and 400,000 images in full and weak annotations, respectively. This competition aims to explore the abilities of state-of-the-art methods to detect and recognize text instances from large-scale street view images, closing gaps between research benchmarks and real applications. During the competition period, a total number of 41 teams participate in the two tasks with 132 valid submissions, i.e., text detection and end-to-end text spotting. This paper includes dataset descriptions, task definitions, evaluation protocols and results summaries of ICDAR 2019-LSVT challenge.

Chinese scene text reading is one of the most challenging problems in computer vision and has attracted great interest. Different from English text, Chinese has more than 6000 commonly used characters and Chinese characters can be arranged in various layouts with numerous fonts. The Chinese signboards in street view are a good choice for Chinese scene text images since they have different backgrounds, fonts and layouts. We organized a competition called ICDAR2019-ReCTS, which mainly focuses on reading Chinese text on signboard. This report presents the final results of the competition. A large-scale dataset of 25,000 annotated signboard images, in which all the text lines and characters are annotated with locations and transcriptions, were released. Four tasks, namely character recognition, text line recognition, text line detection and end-to-end recognition were set up. Besides, considering the Chinese text ambiguity issue, we proposed a multi ground truth (multi-GT) evaluation method to make evaluation fairer. The competition started on March 1, 2019 and ended on April 30, 2019. 262 submissions from 46 teams are received. Most of the participants come from universities, research institutes, and tech companies in China. There are also some participants from the United States, Australia, Singapore, and Korea. 21 teams submit results for Task 1, 23 teams submit results for Task 2, 24 teams submit results for Task 3, and 13 teams submit results for Task 4. The official website for the competition is http://rrc.cvc.uab.es/?ch=12.

Cross-modal retrieval methods have been significantly improved in last years with the use of deep neural networks and large-scale annotated datasets such as ImageNet and Places. However, collecting and annotating such datasets requires a tremendous amount of human effort and, besides, their annotations are limited to discrete sets of popular visual classes that may not be representative of the richer semantics found on large-scale cross-modal retrieval datasets. In this paper, we present a self-supervised cross-modal retrieval framework that leverages as training data the correlations between images and text on the entire set of Wikipedia articles. Our method consists in training a CNN to predict: (1) the semantic context of the article in which an image is more probable to appear as an illustration, and (2) the semantic context of its caption. Our experiments demonstrate that the proposed method is not only capable of learning discriminative visual representations for solving vision tasks like classification, but that the learned representations are better for cross-modal retrieval when compared to supervised pre-training of the network on the ImageNet dataset.

In this paper we present a segmentation-free system for reading text in natural scenes. A CNN architecture is trained in an end-to-end manner, and is able to directly output readings without any explicit text localization step. In order to validate our proposal, we focus on the specific case of reading utility meters. We present our results in a large dataset of images acquired by different users and devices, so text appears in any location, with different sizes, fonts and lengths, and the images present several distortions such as dirt, illumination highlights or blur.

The ICDAR Robust Reading Competition (RRC), initiated in 2003 and re-established in 2011, has become a de-facto evaluation standard for robust reading systems and algorithms. Concurrent with its second incarnation in 2011, a continuous effort started to develop an on-line framework to facilitate the hosting and management of competitions. This paper outlines the Robust Reading Competition Annotation and Evaluation Platform, the backbone of the competitions. The RRC Annotation and Evaluation Platform is a modular framework, fully accessible through on-line interfaces. It comprises a collection of tools and services for managing all processes involved with defining and evaluating a research task, from dataset definition to annotation management, evaluation specification and results analysis. Although the framework has been designed with robust reading research in mind, many of the provided tools are generic by design. All aspects of the RRC Annotation and Evaluation Framework are available for research use.

End-to-end training from scratch of current deep architectures for new computer vision problems would require Imagenet-scale datasets, and this is not always possible. In this paper we present a method that is able to take advantage of freely available multi-modal content to train computer vision algorithms without human supervision. We put forward the idea of performing self-supervised learning of visual features by mining a large scale corpus of multi-modal (text and image) documents. We show that discriminative visual features can be learnt efficiently by training a CNN to predict the semantic context in which a particular image is more probable to appear as an illustration. For this we leverage the hidden semantic structures discovered in the text corpus with a well-known topic modeling technique. Our experiments demonstrate state of the art performance in image classification, object detection, and multi-modal retrieval compared to recent self-supervised or natural-supervised approaches.

In this paper we present the LSDE string representation and its application to handwritten word spotting. LSDE is a novel embedding approach for representing strings that learns a space in which distances between projected points are correlated with the Levenshtein edit distance between the original strings. We show how such a representation produces a more semantically interpretable retrieval from the user's perspective than other state of the art ones such as PHOC and DCToW. We also conduct a preliminary handwritten word spotting experiment on the George Washington dataset.

This report presents the final results of the ICDAR 2017 Robust Reading Challenge on COCO-Text. A challenge on scene text detection and recognition based on the largest real scene text dataset currently available: the COCO-Text dataset. The competition is structured around three tasks: Text Localization, Cropped Word Recognition and End-To-End Recognition. The competition received a total of 27 submissions over the different opened tasks. This report describes the datasets and the ground truth, details the performance evaluation protocols used and presents the final results along with a brief summary of the participating methods.

Results of ICDAR 2017 Robust Reading Challenge on Omnidirectional Video are presented. This competition uses Downtown Osaka Scene Text (DOST) Dataset that was captured in Osaka, Japan with an omnidirectional camera. Hence, it consists of sequential images (videos) of different view angles. Regarding the sequential images as videos (video mode), two tasks of localisation and end-to-end recognition are prepared. Regarding them as a set of still images (still image mode), three tasks of localisation, cropped word recognition and end-to-end recognition are prepared. As the dataset has been captured in Japan, the dataset contains Japanese text but also include text consisting of alphanumeric characters (Latin text). Hence, a submitted result for each task is evaluated in three ways: using Japanese only ground truth (GT), using Latin only GT and using combined GTs of both. Finally, by the submission deadline, we have received two submissions in the text localisation task of the still image mode. We intend to continue the competition in the open mode. Expecting further submissions, in this report we provide baseline results in all the tasks in addition to the submissions from the community.

Text detection and recognition in a natural environment are key components of many applications, ranging from business card digitization to shop indexation in a street. This competition aims at assessing the ability of state-of-the-art methods to detect Multi-Lingual Text (MLT) in scene images, such as in contents gathered from the Internet media and in modern cities where multiple cultures live and communicate together. This competition is an extension of the Robust Reading Competition (RRC) which has been held since 2003 both in ICDAR and in an online context. The proposed competition is presented as a new challenge of the RRC. The dataset built for this challenge largely extends the previous RRC editions in many aspects: the multi-lingual text, the size of the dataset, the multi-oriented text, the wide variety of scenes. The dataset is comprised of 18,000 images which contain text belonging to 9 languages. The challenge is comprised of three tasks related to text detection and script classification. We have received a total of 16 participations from the research and industrial communities. This paper presents the dataset, the tasks and the findings of this RRC-MLT challenge.

Hundreds of millions of figures are available in the biomedical literature, representing important biomedical experimental evidence. Since text is a rich source of information in figures, automatically extracting such text may assist in the task of mining figure information and understanding biomedical documents. Unlike images in the open domain, biomedical figures present a variety of unique challenges. For example, biomedical figures typically have complex layouts, small font sizes, short text, specific text, complex symbols and irregular text arrangements. This paper presents the final results of the ICDAR 2017 Competition on Text Extraction from Biomedical Literature Figures (ICDAR2017 DeTEXT Competition), which aims at extracting (detecting and recognizing) text from biomedical literature figures. Similar to text extraction from scene images and web pictures, ICDAR2017 DeTEXT Competition includes three major tasks, i.e., text detection, cropped word recognition and end-to-end text recognition. Here, we describe in detail the data set, tasks, evaluation protocols and participants of this competition, and report the performance of the participating methods.

This paper focuses on the problem of script identification in unconstrained scenarios. Script identification is an important prerequisite to recognition, and an indispensable condition for automatic text understanding systems designed for multi-language environments. Although widely studied for document images and handwritten documents, it remains an almost unexplored territory for scene text images. We detail a novel method for script identification in natural images that combines convolutional features and the Naive-Bayes Nearest Neighbor classifier. The proposed framework efficiently exploits the discriminative power of small stroke-parts, in a fine-grained classification framework. In addition, we propose a new public benchmark dataset for the evaluation of joint text detection and script identification in natural scenes. Experiments done in this new dataset demonstrate that the proposed method yields state of the art results, while it generalizes well to different datasets and variable number of scripts. The evidence provided shows that multi-lingual scene text recognition in the wild is a viable proposition. Source code of the proposed method is made available online.

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.

In this paper we introduce a script identification method based on hand-crafted texture features and an artificial neural network. The proposed pipeline achieves near state-of-the-art performance for script identification of video-text and state-of-the-art performance on visual language identification of handwritten text. More than using the deep network as a classifier, the use of its intermediary activations as a learned metric demonstrates remarkable results and allows the use of discriminative models on unknown classes. Comparative experiments in video-text and text in the wild datasets provide insights on the internals of the proposed deep network.

Object Proposals is a recent computer vision technique receiving increasing interest from the research community. Its main objective is to generate a relatively small set of bounding box proposals that are most likely to contain objects of interest. The use of Object Proposals techniques in the scene text understanding field is innovative. Motivated by the success of powerful while expensive techniques to recognize words in a holistic way, Object Proposals techniques emerge as an alternative to the traditional text detectors. In this paper we study to what extent the existing generic Object Proposals methods may be useful for scene text understanding. Also, we propose a new Object Proposals algorithm that is specifically designed for text and compare it with other generic methods in the state of the art. Experiments show that our proposal is superior in its ability of producing good quality word proposals in an efficient way. The source code of our method is made publicly available1.

Spatial verification is typically employed to check the spatial consistency among matched local features and to remove outliers. However, when looking for multiple instances of the query within a target image, RANSAC algorithms which are widely applied in many one-to-one matching applications might fail due to the large proportion of “outliers” - correct matches corresponding to other instances. On the other hand, geometrical verification methods are more robust to outliers but usually suffer from high computational costs. In this paper, we introduce a novel two-step line verification method which is more flexible than existing methods and leads to lower computational complexity especially when multiple instances of a query are sought. We study this approach within an information extraction scenario, where the objective is to locate document structures indicative of certain type of information (e.g. different records on invoices).

This paper deals with Query By String word spotting in scene images. A hierarchical text segmentation algorithm based on text specific selective search is used to find text regions. These regions are indexed per character n-grams present in the text region. An attribute representation based on Pyramidal Histogram of Characters (PHOC) is used to compare text regions with the query text. For generation of the index a similar attribute space based Pyramidal Histogram of character n-grams is used. These attribute models are learned using linear SVMs over the Fisher Vector [1] representation of the images along with the PHOC labels of the corresponding strings.

Results of the ICDAR 2015 Robust Reading Competition are presented. A new Challenge 4 on Incidental Scene Text has been added to the Challenges on Born-Digital Images, Focused Scene Images and Video Text. Challenge 4 is run on a newly acquired dataset of 1,670 images evaluating Text Localisation, Word Recognition and End-to-End pipelines. In addition, the dataset for Challenge 3 on Video Text has been substantially updated with more video sequences and more accurate ground truth data. Finally, tasks assessing End-to-End system performance have been introduced to all Challenges. The competition took place in the first quarter of 2015, and received a total of 44 submissions. Only the tasks newly introduced in 2015 are reported on. The datasets, the ground truth specification and the evaluation protocols are presented together with the results and a brief summary of the participating methods.

Sampling Local Binary Patterns, a variant of Local Binary Patterns (LBP) for text-as-texture classification. By adapting and extending the standard LBP operator to the particularities of text we get a generic text-as-texture classification scheme and apply it to writer identification. In experiments on CVL and ICDAR 2013 datasets, the proposed feature-set and a simple end-to-end pipeline demonstrate State-Of-the-Art (SOA) performance. Among the SOA, the proposed method is the only one that is based on dense extraction of a single local feature descriptor. This makes it fast and applicable at the earliest stages in a DIA pipeline without the need for segmentation, binarization, or extraction of multiple features.

With increasing computational power, the trend in unconstrained text recognition is going towards whole document processing. For this task, more sophisticated language models can be employed. One approach is to take advantage the fact that the text of a document normally deals with a specific topic and hence the word occurrence probability is biased. Cache language models combine information about recent words, the cache, with a general statistical language model to increase the recognition rate. In this work we introduce a modified version of the cache language model to the task of handwriting recognition, where the N-best recognition output of the entire document is used to refine the language model for a consecutive recognition pass. An experimental evaluation on the IAM database demonstrates that the word error rate can be reduced with the proposed cache language model.

This work presents a set of on-line software tools for creating ground truth and calculating performance evaluation metrics for text extraction tasks such as localization, segmentation and recognition. The platform supports the definition of comprehensive ground truth information at different text representation levels while it offers centralised management and quality control of the ground truthing effort. It implements a range of state of the art performance evaluation algorithms and offers functionality for the definition of evaluation scenarios, on-line calculation of various performance metrics and visualisation of the results. The presented platform, which comprises the backbone of the ICDAR 2011 (challenge 1) and 2013 (challenges 1 and 2) Robust Reading competitions, is now made available for public use.

Human detection in computer vision field is an active field of research. Extending this to human-like drawings such as the main characters in comic book stories is not trivial. Comics analysis is a very recent field of research at the intersection of graphics, texts, objects and people recognition. The detection of the main comic characters is an essential step towards a fully automatic comic book understanding. This paper presents a color-based approach for comics character retrieval using content-based drawing retrieval and color palette.

Since the document structure carries valuable discriminative information, plenty of efforts have been made for extracting and understanding document structure among which layout analysis approaches are the most commonly used. In this paper, Distance Transform based MSER (DTMSER) is employed to efficiently extract the document structure as a dendrogram of key-regions which roughly correspond to structural elements such as characters, words and paragraphs. Inspired by the Bag of Words (BoW) framework, we propose an efficient method for structural document matching by representing the document image as a histogram of key-region pairs encoding structural relationships. Applied to the scenario of document image retrieval, experimental results demonstrate a remarkable improvement when comparing the proposed method with typical BoW and pyramidal BoW methods.

We present a hybrid algorithm for detection and tracking of text in natural scenes that goes beyond the full-detection approaches in terms of time performance optimization. A state-of-the-art scene text detection module based on Maximally Stable Extremal Regions (MSER) is used to detect text asynchronously, while on a separate thread detected text objects are tracked by MSER propagation. The cooperation of these two modules yields real time video processing at high frame rates even on low-resource devices.

In this paper we argue that a key-region detector designed to take into account the special characteristics of document images can result in the detection of less and more meaningful key-regions. We propose a fast key-region detector able to capture aspects of the structural information of the document, and demonstrate its efficiency by comparing against standard detectors in an administrative document retrieval scenario. We show that using the proposed detector results to a smaller number of detected key-regions and higher performance without any drop in speed compared to standard state of the art detectors.

Scene text extraction methodologies are usually based in classification of individual regions or patches, using a priori knowledge for a given script or language. Human perception of text, on the other hand, is based on perceptual organisation through which text emerges as a perceptually significant group of atomic objects. Therefore humans are able to detect text even in languages and scripts never seen before. In this paper, we argue that the text extraction problem could be posed as the detection of meaningful groups of regions. We present a method built around a perceptual organisation framework that exploits collaboration of proximity and similarity laws to create text-group hypotheses. Experiments demonstrate that our algorithm is competitive with state of the art approaches on a standard dataset covering text in variable orientations and two languages.

In this paper we present a method for the segmentation of continuous page streams into multipage documents and the simultaneous classification of the resulting documents. We first present an approach to combine the multiple pages of a document into a single feature vector that represents the whole document. Despite its simplicity and low computational cost, the proposed representation yields results comparable to more complex methods in multipage document classification tasks. We then exploit this representation in the context of page stream segmentation. The most plausible segmentation of a page stream into a sequence of multipage documents is obtained by optimizing a statistical model that represents the probability of each segmented multipage document belonging to a particular class. Experimental results are reported on a large sample of real administrative multipage documents.

This report presents the final results of the ICDAR 2013 Robust Reading Competition. The competition is structured in three Challenges addressing text extraction in different application domains, namely born-digital images, real scene images and real-scene videos. The Challenges are organised around specific tasks covering text localisation, text segmentation and word recognition. The competition took place in the first quarter of 2013, and received a total of 42 submissions over the different tasks offered. This report describes the datasets and ground truth specification, details the performance evaluation protocols used and presents the final results along with a brief summary of the participating methods.

Comic books constitute an important cultural heritage asset in many countries. Digitization combined with subsequent comic book understanding would enable a variety of new applications, including content-based retrieval and content retargeting. Document understanding in this domain is challenging as comics are semi-structured documents, combining semantically important graphical and textual parts. Few studies have been done in this direction. In this work we detail a novel approach for closed and non-closed speech balloon localization in scanned comic book pages, an essential step towards a fully automatic comic book understanding. The approach is compared with existing methods for closed balloon localization found in the literature and results are presented.

We propose a method to acquire multispectral data with handheld devices with front-mounted RGB cameras. We propose to use the display of the device as an illuminant while the camera captures images illuminated by the red, green and blue primaries of the display. Three illuminants and three response functions of the camera lead to nine response values which are used for reflectance estimation. Results are promising and show that the accuracy of the spectral reconstruction improves in the range of 30-40% over the spectral reconstruction based on a single illuminant. Furthermore, we propose to compute sensor-illuminant aware linear basis by discarding the part of the reflectances that falls in the sensor-illuminant null-space. We show experimentally that optimizing reflectance estimation on these new basis functions decreases the RMSE significantly over basis functions that are independent to sensor-illuminant. We conclude that, multispectral data acquisition is potentially possible with consumer hand-held devices such as tablets, mobiles, and laptops, opening up applications which are currently considered to be unrealistic.

This paper presents the results of the first Challenge of ICDAR 2011 Robust Reading Competition. Challenge 1 is focused on the extraction of text from born-digital images, specifically from images found in Web pages and emails. The challenge was organized in terms of three tasks that look at different stages of the process: text localization, text segmentation and word recognition. In this paper we present the results of the challenge for all three tasks, and make an open call for continuous participation outside the context of ICDAR 2011.

The availability of open, ground-truthed datasets and clear performance metrics is a crucial factor in the development of an application domain. The domain of colour text image analysis (real scenes, Web and spam images, scanned colour documents) has traditionally suffered from a lack of a comprehensive performance evaluation framework. Such a framework is extremely difficult to specify, and corresponding pixel-level accurate information tedious to define. In this paper we discuss the challenges and technical issues associated with developing such a framework. Then, we describe a complete framework for the evaluation of text extraction methods at multiple levels, provide a detailed ground-truth specification and present a case study on how this framework can be used in a real-life situation.

In this paper, we propose a novel rotation and scale invariant method for colour logo retrieval and classification, which involves performing a simple colour segmentation and subsequently describing each of the resultant colour components based on a set of topological and colour features. A polar representation is used to represent the logo and the subsequent logo matching is based on Cyclic Dynamic Time Warping (CDTW). We also show how combining information about the global distribution of the logo components and their local neighbourhood using the Delaunay triangulation allows to improve the results. All experiments are performed on a dataset of 2500 instances of 100 colour logo images in different rotations and scales.

In this paper we present a method for the retrieval of images in terms of perceptual similarity. Local color information is added to the shape context descriptor in order to obtain an object description integrating both shape and color as visual cues. We use a color naming algorithm in order to represent the color information from a perceptual point of view. The proposed method has been tested in two different applications, an object retrieval scenario based on color sketch queries and a color trademark retrieval problem. Experimental results show that the addition of the color information significantly outperforms the sole use of the shape context descriptor.

The extraction of textual content from colour documents of a graphical nature is a complicated task. The text can be rendered in any colour, size and orientation while the existence of complex background graphics with repetitive patterns can make its localization and segmentation extremely difficult. Here, we propose a new method for extracting textual content from such colour images that makes no assumption as to the size of the characters, their orientation or colour, while it is tolerant to characters that do not follow a straight baseline. We evaluate this method on a collection of documents with historical connotations: the posters from the Spanish Civil War.

The use of gradients in text images is nowadays quite frequent. Existing segmentation methods encounter serious problems when it comes to modern text images where gradients might appear in the background or the foreground or both at the same time. This paper presents an approach for lightness gradient areas detection based on the Hough Transform. The issues arising are discussed, and results are presented on a dataset comprising Web images, logos and scanned documents.

This article describes a sketch-based framework for semi-automatic annotation of historical document collections. It is motivated by the fact that fully automatic methods, while helpful for extracting metadata from large collections, have two main drawbacks in a real-world application: (i) they are error-prone and (ii) they only capture a subset of all the knowledge in the document base, both meaning that manual intervention is always required. Therefore, we have developed a practical framework for allowing experts to extract knowledge from document collections in a sketch-based scenario. The main possibilities of the proposed framework are: (a) browsing the collection efficiently, (b) providing gestures for metadata input, (c) supporting handwritten notes and (d) providing gestures for launching automatic extraction processes such as OCR or word spotting.

The vignette effect (radial fall-off) is commonly encountered in images obtained through certain image acquisition setups and can seriously hinder automatic analysis processes. In this paper we present a fast and efficient method for dealing with vignetting in the context of object segmentation in an existing industrial inspection setup. The vignette effect is modelled here as a circular, non-linear gradient. The method estimates the gradient parameters and employs them to perform segmentation. Segmentation results on a variety of images indicate that the presented method is able to successfully tackle the vignette effect.

This paper presents a flexible approach to extracting content from scanned historical documents using semantic information. The final electronic document is the result of a "digital historical document lifecycle" process, where the expert knowledge of the historian/archivist user is incorporated at different stages. Results show that such a conversion strategy aided by (expert) user-specified semantic information and which enables the processing of individual parts of the document in a specialised way, produces superior (in a variety of significant ways) results than document analysis and understanding techniques devised for contemporary documents.

This paper describes the lifecycle of a digital historical document, from template-based structure definition through to content extraction from the scanned pages and its final reconstitution as an electronic document (combining content and semantic information) along with the tools that have been created to realise each stage in the lifecycle. The whole approach is described in the context of different types of typewritten documents relating to prisoners in World-War II concentration camps and is the result of a multinational collaboration under the MEMORIAL project funded (€1.5M) by the European Union (www.memorial-project.info). Extensive tests with historians/archivists and evaluation of the content extraction results indicate the superior performance of the whole semantics-driven approach both over manual transcription and over the semi-automated application of off-the-shelf OCR and the use of a conventional (text and layout) document format.

There is a significant need to recognise the text in images on Web pages, both for effective indexing and for presentation by non-visual means (e.g., audio). This paper presents and compares two novel methods for the segmentation of characters for subsequent extraction and recognition. The novelty of both approaches is the combination of (different in each case) topological features of characters with an anthropocentric perspective of colour perception - in preference to RGB space analysis. Both approaches enable the extraction of text in complex situations such as in the presence of varying colour and texture (characters and background).