Abstract: 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

Abstract: This paper introduces the DocILE benchmark with the largest dataset of business documents for the tasks of Key Information Localization and Extraction and Line Item Recognition. It contains 6.7k annotated business documents, 100k synthetically generated documents, and nearly 1M unlabeled documents for unsupervised pre-training. The dataset has been built with knowledge of domain- and task-specific aspects, resulting in the following key features: (i) annotations in 55 classes, which surpasses the granularity of previously published key information extraction datasets by a large margin; (ii) Line Item Recognition represents a highly practical information extraction task, where key information has to be assigned to items in a table; (iii) documents come from numerous layouts and the test set includes zero- and few-shot cases as well as layouts commonly seen in the training set. The benchmark comes with several baselines, including RoBERTa, LayoutLMv3 and DETR-based Table Transformer; applied to both tasks of the DocILE benchmark, with results shared in this paper, offering a quick starting point for future work. The dataset, baselines and supplementary material are available at https://github.com/rossumai/docile.

Abstract: A method for estimating the non-linear gamma transfer function of liquid–crystal displays (LCDs) without the need of a photometric measurement device was described by Xiao et al. (2011) [1]. It relies on observer’s judgments of visual luminance by presenting eight half-tone patterns with luminances from 1/9 to 8/9 of the maximum value of each colour channel. These half-tone patterns were distributed over the screen both over the vertical and horizontal viewing axes. We conducted a series of photometric and psychophysical measurements (consisting in the simultaneous presentation of half-tone patterns in each trial) to evaluate whether the angular dependency of the light generated by three different LCD technologies would bias the results of these gamma transfer function estimations. Our results show that there are significant differences between the gamma transfer functions measured and produced by observers at different viewing angles. We suggest appropriate modifications to the Xiao et al. paradigm to counterbalance these artefacts which also have the advantage of shortening the amount of time spent in collecting the psychophysical measurements.

Abstract: We present ECOC-DRF, a framework where potential functions for Discriminative Random Fields are formulated as an ensemble of classifiers. We introduce the label trick, a technique to express transitions in the pairwise potential as meta-classes. This allows to independently learn any possible transition between labels without assuming any pre-defined model. The Error Correcting Output Codes matrix is used as ensemble framework for the combination of margin classifiers. We apply ECOC-DRF to a large set of classification problems, covering synthetic, natural and medical images for binary and multi-class cases, outperforming state-of-the art in almost all the experiments.

Abstract: This chapter describes the most relevant approaches for the analysis of graphical documents. The graphics recognition pipeline can be splitted into three tasks. The low level or lexical task extracts the basic units composing the document. The syntactic level is focused on the structure, i.e., how graphical entities are constructed, and involves the location and classification of the symbols present in the document. The third level is a functional or semantic level, i.e., it models what the graphical symbols do and what they mean in the context where they appear. This chapter covers the lexical level, while the next two chapters are devoted to the syntactic and semantic level, respectively. The main problems reviewed in this chapter are raster-to-vector conversion (vectorization algorithms) and the separation of text and graphics components. The research and industrial communities have provided standard methods achieving reasonable performance levels. Hence, graphics recognition techniques can be considered to be in a mature state from a scientific point of view. Additionally this chapter provides insights on some related problems, namely, the extraction and recognition of dimensions in engineering drawings, and the recognition of hatched and tiled patterns. Both problems are usually associated, even integrated, in the vectorization process.

Abstract: Advanced segmentation techniques in the surveillance domain deal with shadows to avoid distortions when detecting moving objects. Most approaches for shadow detection are still typically restricted to penumbra shadows and cannot cope well with umbra shadows. Consequently, umbra shadow regions are usually detected as part of moving objects, thus aecting the performance of the nal detection. In this paper we address the detection of both penumbra and umbra shadow regions. First, a novel bottom-up approach is presented based on gradient and colour models, which successfully discriminates between chromatic moving cast shadow regions and those regions detected as moving objects. In essence, those regions corresponding to potential shadows are detected based on edge partitioning and colour statistics. Subsequently (i) temporal similarities between textures and (ii) spatial similarities between chrominance angle and brightness distortions are analysed for each potential shadow region for detecting the umbra shadow regions. Our second contribution renes even further the segmentation results: a tracking-based top-down approach increases the performance of our bottom-up chromatic shadow detection algorithm by properly correcting non-detected shadows.
To do so, a combination of motion lters in a data association framework exploits the temporal consistency between objects and shadows to increase
the shadow detection rate. Experimental results exceed current state-of-the-
art in shadow accuracy for multiple well-known surveillance image databases which contain dierent shadowed materials and illumination conditions.

Abstract: An efficient snake formulation should conform to the idea of picking the smoothest curve among all the shapes approximating an object of interest. In current geodesic snakes, the regularizing curvature also affects the convergence stage, hindering the latter at concave regions. In the present work, we make use of characteristic functions to define a novel geodesic formulation that decouples regularity and convergence. This term decoupling endows the snake with higher adaptability to non-convex shapes. Convergence is ensured by splitting the definition of the external force into an attractive vector field and a repulsive one. In our paper, we propose to use likelihood maps as approximation of characteristic functions of object appearance. The better efficiency and accuracy of our decoupled scheme are illustrated in the particular case of feature space-based segmentation.

Abstract: Deep learning has provided promising results in various applications; however, algorithms tend to be overconfident in their predictions, even though they may be entirely wrong. Particularly for critical applications, the model should provide answers only when it is very sure of them. This article presents a Bayesian version of two different state-of-the-art semantic segmentation methods to perform multi-class segmentation of foods and estimate the uncertainty about the given predictions. The proposed methods were evaluated on three public pixel-annotated food datasets. As a result, we can conclude that Bayesian methods improve the performance achieved by the baseline architectures and, in addition, provide information to improve decision-making. Furthermore, based on the extracted uncertainty map, we proposed three measures to rank the images according to the degree of noisy annotations they contained. Note that the top 135 images ranked by one of these measures include more than half of the worst-labeled food images.