Abstract: Document Visual Question Answering (DocVQA) refers to the task of answering questions from document images. Existing work on DocVQA only considers single-page documents. However, in real scenarios documents are mostly composed of multiple pages that should be processed altogether. In this work we extend DocVQA to the multi-page scenario. For that, we first create a new dataset, MP-DocVQA, where questions are posed over multi-page documents instead of single pages. Second, we propose a new hierarchical method, Hi-VT5, based on the T5 architecture, that overcomes the limitations of current methods to process long multi-page documents. The proposed method is based on a hierarchical transformer architecture where the encoder summarizes the most relevant information of every page and then, the decoder takes this summarized information to generate the final answer. Through extensive experimentation, we demonstrate that our method is able, in a single stage, to answer the questions and provide the page that contains the relevant information to find the answer, which can be used as a kind of explainability measure.

Abstract: Multimodal learning from document data has achieved great success lately as it allows to pre-train semantically meaningful features as a prior into a learnable downstream approach. In this paper, we approach the document classification problem by learning cross-modal representations through language and vision cues, considering intra- and inter-modality relationships. Instead of merging features from different modalities into a common representation space, the proposed method exploits high-level interactions and learns relevant semantic information from effective attention flows within and across modalities. The proposed learning objective is devised between intra- and inter-modality alignment tasks, where the similarity distribution per task is computed by contracting positive sample pairs while simultaneously contrasting negative ones in the common feature representation space}. Extensive experiments on public document classification datasets demonstrate the effectiveness and the generalization capacity of our model on both low-scale and large-scale datasets.

Abstract: The advent of recurrent neural networks for handwriting recognition marked an important milestone reaching impressive recognition accuracies despite the great variability that we observe across different writing styles. Sequential architectures are a perfect fit to model text lines, not only because of the inherent temporal aspect of text, but also to learn probability distributions over sequences of characters and words. However, using such recurrent paradigms comes at a cost at training stage, since their sequential pipelines prevent parallelization. In this work, we introduce a non-recurrent approach to recognize handwritten text by the use of transformer models. We propose a novel method that bypasses any recurrence. By using multi-head self-attention layers both at the visual and textual stages, we are able to tackle character recognition as well as to learn language-related dependencies of the character sequences to be decoded. Our model is unconstrained to any predefined vocabulary, being able to recognize out-of-vocabulary words, i.e. words that do not appear in the training vocabulary. We significantly advance over prior art and demonstrate that satisfactory recognition accuracies are yielded even in few-shot learning scenarios.

Abstract: Handwritten document images can be highly affected by degradation for different reasons: Paper ageing, daily-life scenarios (wrinkles, dust, etc.), bad scanning process and so on. These artifacts raise many readability issues for current Handwritten Text Recognition (HTR) algorithms and severely devalue their efficiency. In this paper, we propose an end to end architecture based on Generative Adversarial Networks (GANs) to recover the degraded documents into a and form. Unlike the most well-known document binarization methods, which try to improve the visual quality of the degraded document, the proposed architecture integrates a handwritten text recognizer that promotes the generated document image to be more readable. To the best of our knowledge, this is the first work to use the text information while binarizing handwritten documents. Extensive experiments conducted on degraded Arabic and Latin handwritten documents demonstrate the usefulness of integrating the recognizer within the GAN architecture, which improves both the visual quality and the readability of the degraded document images. Moreover, we outperform the state of the art in H-DIBCO challenges, after fine tuning our pre-trained model with synthetically degraded Latin handwritten images, on this task.

Abstract: Sequence-to-sequence models have recently become very popular for tackling
handwritten word recognition problems. However, how to effectively integrate an external language model into such recognizer is still a challenging
problem. The main challenge faced when training a language model is to
deal with the language model corpus which is usually different to the one
used for training the handwritten word recognition system. Thus, the bias
between both word corpora leads to incorrectness on the transcriptions, providing similar or even worse performances on the recognition task. In this
work, we introduce Candidate Fusion, a novel way to integrate an external
language model to a sequence-to-sequence architecture. Moreover, it provides suggestions from an external language knowledge, as a new input to
the sequence-to-sequence recognizer. Hence, Candidate Fusion provides two
improvements. On the one hand, the sequence-to-sequence recognizer has
the flexibility not only to combine the information from itself and the language model, but also to choose the importance of the information provided
by the language model. On the other hand, the external language model
has the ability to adapt itself to the training corpus and even learn the
most commonly errors produced from the recognizer. Finally, by conducting
comprehensive experiments, the Candidate Fusion proves to outperform the
state-of-the-art language models for handwritten word recognition tasks.

Abstract: In this paper we propose to embed SMPL within a deep-based model to accurately estimate 3D pose and shape from a still RGB image. We use CNN-based 3D joint predictions as an intermediate representation to regress SMPL pose and shape parameters. Later, 3D joints are reconstructed again in the SMPL output. This module can be seen as an autoencoder where the encoder is a deep neural network and the decoder is SMPL model. We refer to this as SMPL reverse (SMPLR). By implementing SMPLR as an encoder-decoder we avoid the need of complex constraints on pose and shape. Furthermore, given that in-the-wild datasets usually lack accurate 3D annotations, it is desirable to lift 2D joints to 3D without pairing 3D annotations with RGB images. Therefore, we also propose a denoising autoencoder (DAE) module between CNN and SMPLR, able to lift 2D joints to 3D and partially recover from structured error. We evaluate our method on SURREAL and Human3.6M datasets, showing improvement over SMPL-based state-of-the-art alternatives by about 4 and 12 mm, respectively.

Abstract: The aim of this paper is to revisit an old theory of texture perception and
update its computational implementation by extending it to colour. With this in mind we try to capture the optimality of perceptual systems. This is achieved in the proposed approach by sharing well-known early stages of the visual processes and extracting low-dimensional features that perfectly encode adequate properties for a large variety of textures without needing further learning stages. We propose several descriptors in a bag-of-words framework that are derived from different quantisation models on to the feature spaces. Our perceptual features are directly given by the shape and colour attributes of image blobs, which are the textons. In this way we avoid learning visual words and directly build the vocabularies on these lowdimensionaltexton spaces. Main differences between proposed descriptors rely on how co-occurrence of blob attributes is represented in the vocabularies. Our approach overcomes current state-of-art in colour texture description which is proved in several experiments on large texture datasets.

Abstract: There are printed complex documents where text lines of a single page may have different orientations or the text lines may be curved in shape. As a result, it is difficult to detect the skew of such documents and hence character segmentation and recognition of such documents are a complex task. In this paper, using background and foreground information we propose a novel scheme towards the recognition of Indian complex documents of Bangla and Devnagari script. In Bangla and Devnagari documents usually characters in a word touch and they form cavity regions. To take care of these cavity regions, background information of such documents is used. Convex hull and water reservoir principle have been applied for this purpose. Here, at first, the characters are segmented from the documents using the background information of the text. Next, individual characters are recognized using rotation invariant features obtained from the foreground part of the characters.
For character segmentation, at first, writing mode of a touching component (word) is detected using water reservoir principle based features. Next, depending on writing mode and the reservoir base-region of the touching component, a set of candidate envelope points is then selected from the contour points of the component. Based on these candidate points, the touching component is finally segmented into individual characters. For recognition of multi-sized/multi-oriented characters the features are computed from different angular information obtained from the external and internal contour pixels of the characters. These angular information are computed in such a way that they do not depend on the size and rotation of the characters. Circular and convex hull rings have been used to divide a character into smaller zones to get zone-wise features for higher recognition results. We combine circular and convex hull features to improve the results and these features are fed to support vector machines (SVM) for recognition. From our experiment we obtained recognition results of 99.18% (98.86%) accuracy when tested on 7515 (7874) Devnagari (Bangla) characters.