Abstract: This paper presents a new method for human detection based on a multiresolution cascade of Histograms of Oriented Gradients (HOG) that can highly reduce the computational cost of the detection search without affecting accuracy. The method consists of a cascade of sliding window detectors. Each detector is a Support Vector Machine (SVM) composed by features at different resolution, from coarse for the first level to fine for the last one.
Considering that the spatial stride of the sliding window search is affected by the HOG features size, unlike previous methods based on Adaboost cascades, we can adopt a spatial stride inversely proportional to the features resolution. This produces that the speed-up of the cascade is not only due to the low number of features that need to be computed in the first levels, but also to the lower number of detection windows that needs to be evaluated.
Experimental results shows that our method permits a detection rate comparable with the state of the art, but at the same time a gain in the speed of the detection search of 10-20 times depending on the cascade configuration.

Abstract: Despite being very successful within the pattern recognition and machine learning community, graph-based methods are often unusable because of the lack of mathematical operations defined in graph domain. Graph embedding, which maps graphs to a vectorial space, has been proposed as a way to tackle these difficulties enabling the use of standard machine learning techniques. However, it is well known that graph embedding functions usually suffer from the loss of structural information. In this paper, we consider the hierarchical structure of a graph as a way to mitigate this loss of information. The hierarchical structure is constructed by topologically clustering the graph nodes and considering each cluster as a node in the upper hierarchical level. Once this hierarchical structure is constructed, we consider several configurations to define the mapping into a vector space given a classical graph embedding, in particular, we propose to make use of the stochastic graphlet embedding (SGE). Broadly speaking, SGE produces a distribution of uniformly sampled low-to-high-order graphlets as a way to embed graphs into the vector space. In what follows, the coarse-to-fine structure of a graph hierarchy and the statistics fetched by the SGE complements each other and includes important structural information with varied contexts. Altogether, these two techniques substantially cope with the usual information loss involved in graph embedding techniques, obtaining a more robust graph representation. This fact has been corroborated through a detailed experimental evaluation on various benchmark graph datasets, where we outperform the state-of-the-art methods.

Abstract: Part detection is an important aspect of object recognition. Most approaches apply object proposals to generate hundreds of possible part bounding box candidates which are then evaluated by part classifiers. Recently several methods have investigated directly regressing to a limited set of bounding boxes from deep neural network representation. However, for object parts such methods may be unfeasible due to their relatively small size with respect to the image. We propose a hierarchical method for object and part detection. In a single network we first detect the object and then regress to part location proposals based only on the feature representation inside the object. Experiments show that our hierarchical approach outperforms a network which directly regresses the part locations. We also show that our approach obtains part detection accuracy comparable or better than state-of-the-art on the CUB-200 bird and Fashionista clothing item datasets with only a fraction of the number of part proposals.

Abstract: In this paper, we propose an On-line Appearance-Based Tracker (OABT) for simultaneous tracking of 3D head pose, lips, eyebrows, eyelids and irises in monocular video sequences. In contrast to previously proposed tracking approaches, which deal with face and gaze tracking separately, our OABT can also be used for eyelid and iris tracking, as well as 3D head pose, lips and eyebrows facial actions tracking. Furthermore, our approach applies an on-line learning of changes in the appearance of the tracked target. Hence, the prior training of appearance models, which usually requires a large amount of labeled facial images, is avoided. Moreover, the proposed method is built upon a hierarchical combination of three OABTs, which are optimized using a Levenberg–Marquardt Algorithm (LMA) enhanced with line-search procedures. This, in turn, makes the proposed method robust to changes in lighting conditions, occlusions and translucent textures, as evidenced by our experiments. Finally, the proposed method achieves head and facial actions tracking in real-time.

Abstract: The ability to automatically detect and recognize objects in unconstrained images is becoming more and more critical: from security systems and autonomous robots, to smart phones and augmented reality, intelligent devices need to understand the meaning of images as a composition of semantic objects. This Thesis tackles the problem of fast object detection based on template models. Detection consists of searching for an object in an image by evaluating the similarity between a template model and an image region at each possible location and scale. In this work, we show that using a template model representation based on a multiple resolution hierarchy is an optimal choice that can lead to excellent detection accuracy and fast computation. We implement two different approaches that make use of a hierarchy of multiresolution models: a multiresolution cascade and a coarse-to-fine search. Also, we extend the coarse-to-fine search by introducing a deformable part-based model that achieves state-of-the-art results together with a very reduced computational cost. Finally, we specialize our approach to the challenging task of pedestrian detection from moving vehicles and show that the overall quality of the system outperforms previous works in terms of speed and accuracy.

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.