Vehicle recognition is a challenging task with many useful applications. State-of-the-art methods usually learn discriminative classifiers for different vehicle categories or different viewpoint angles, but little work has explored vehicle recognition using semantic visual attributes. In this paper, we propose a novel iterative multiple instance learning method to model local attributes and viewpoint angles together in the same framework. We expand the standard MISVM formulation to incorporate pairwise constraints based on viewpoint relations within positive exemplars. We show that our method is able to generate discriminative and semantic local attributes for vehicle categories. We also show that we can estimate viewpoint labels more accurately than baselines when these annotations are not available in the training set. We test the technique on the Stanford cars and INRIA vehicles datasets, and compare with other methods.

Most works on image retrieval from text queries have addressed the problem of retrieving semantically relevant images. However, the ability to assess the aesthetic quality of an image is an increasingly important differentiating factor for search engines. In this work, given a semantic query, we are interested in retrieving images which are semantically relevant and score highly in terms of aesthetics/visual quality. We use large-margin classifiers and rankers to learn statistical models capable of ordering images based on the aesthetic and semantic information. In particular, we compare two families of approaches: while the first one attempts to learn a single ranker which takes into account both semantic and aesthetic information, the second one learns separate semantic and aesthetic models. We carry out a quantitative and qualitative evaluation on a recentlypublished large-scale dataset and we show that the second family of techniques significantly outperforms the first one.

With the ever-expanding volume of visual content available, the ability to organize and navigate such content by aesthetic preference is becoming increasingly important. While still in its nascent stage, research into computational models of aesthetic preference already shows great potential. However, to advance research, realistic, diverse and challenging databases are needed. To this end, we introduce a new large-scale database for conducting Aesthetic Visual Analysis: AVA. It contains over 250,000 images along with a rich variety of meta-data including a large number of aesthetic scores for each image, semantic labels for over 60 categories as well as labels related to photographic style. We show the advantages of AVA with respect to existing databases in terms of scale, diversity, and heterogeneity of annotations. We then describe several key insights into aesthetic preference afforded by AVA. Finally, we demonstrate, through three applications, how the large scale of AVA can be leveraged to improve performance on existing preference tasks.

In this paper, we automatically assess the aesthetic properties of images. In the past, this problem has been addressed by hand-crafting features which would correlate with best photographic practices (e.g. “Does this image respect the rule of thirds?”) or with photographic techniques (e.g. “Is this image a macro?”). We depart from this line of research and propose to use generic image descriptors to assess aesthetic quality. We experimentally show that the descriptors we use, which aggregate statistics computed from low-level local features, implicitly encode the aesthetic properties explicitly used by state-of-the-art methods and outperform them by a significant margin.

This paper addresses the problem of font retrieval using a query-by-example paradigm: given a font, retrieve the the most visually similar fonts. We describe a font by (a) rendering a set of reference characters, (b) extracting a feature vector for each reference character and (c) concatenating the-level character descriptors. The similarity between two fonts is simply the similarity between the vectorial representations. Our contribution is an experimental comparison of character-level descriptors of step (b) on a large dataset of 9,000 fonts. The descriptors we chose to evaluate were drawn from the literature on typed and handwritten text analysis. An important conclusion is that the SIFT descriptor, which was shown to be state-of-the-art for object recognition in photographs and for handwriting recognition, yields the best results for font retrieval.

We propose a novel framework for visual saliency detection based on a simple principle: images sharing their global visual appearances are likely to share similar salience. Assuming that an annotated image database is available, we first retrieve the most similar images to the target image; secondly, we build a simple classifier and we use it to generate saliency maps. Finally, we refine the maps and we extract thumbnails. We show that in spite of its simplicity, our framework outperforms state-of-the-art approaches. Another advantage is its ability to deal with visual pop-up and application/task-driven saliency, if appropriately annotated images are available.

In this paper, a multi-camera architecture is presented for heterogeneous targets analysis and tracking. The proposed system can be switched in various configurations enabling automatic and human driven scene analysis. To achieve this, different control strategies have been implemented in order to optimize results in terms of perceptual image quality and pointing precisions.

In this paper a novel integrated multi-camera video-sensor (panoramic scene analysis - PSA) system is proposed for surveillance applications. In the proposed set-up, an omnidirectional imaging device is used in conjunction with a pan, tilt, zoom (PTZ) camera. In particular, the catadioptric sensor is calibrated and used in order to track and classify every single moving object within its 360-degree field of view. Classification is achieved by an innovative algorithm taking advantages from polar image and pedestrian geometrical properties. This approach greatly simplifies classification procedures and enhances target detection and locating system capabilities.

In this paper, techniques and related issues for the definition of a contextual knowledge in ambient-intelligence systems are explored. A logical structure for this kind of system, inspired by a neurobiological brain model, is proposed. Through these considerations, the role and the importance of context awareness in the definition of an artificial organism showing adaptability, pervasiveness, and scalability features are described. Techniques for the definition of a multilayer context representation are explained and practically demonstrated with a test-bed. In the proposed system, a complex event classification is obtained through the fusion of heterogeneous data coming from a set of sensors thanks to the design of a self-organizing map (SOM). The SOM represents the core of the system and testing proofs show good results in the classification of the events taking place in the monitored environment.

The aim of this paper is to present a multi camera system for location estimation inspired to a model inherited from the Data Fusion domain: the Joint Directorate of Laboratories (JDL) model (E. Waltz et al., 1990). The problem specifically faced is the tracking of objects in two complementary applications: intelligent monitoring (video surveillance) and sport viewing (football players tracking), where multiple occluding objects have to be successfully segmented and located using different features such as color, position and dynamics.

In this paper, an algorithm for tracking multiple rigid and non-rigid objects in conditions of occlusion is presented. The proposed method is based on a scalable and adaptive model based on joint information of color and shape. Through a GHT (generalized Hough transform) based voting method the center of mass of each object can be determined in real time with a good degree of precision. Quantitative and qualitative results are presented to validate the efficiency of the method

A system for face detection in outdoor environments for multisensor video surveillance applications is presented. The system is characterized by a combination of two pan-tilt video cameras, which cooperate in order to track and to characterize moving objects with positioning and biometric informations. The final result of the action of the system is the collection of small video shots regarding the face of humans detected in outdoor environments with a robust behavior.

This paper proposes a novel architecture for multisensor data fusion in the context of ambient intelligence (Ami). The proposed system integrates a heterogeneous network of sensors with CCD cameras and computational units working together in a LAN. Activities of humans interacting in the monitored area are detected and classified by combining sensors data output with a neural method.

This paper describes a technique for tracking single objects moving within the guarded scene during dynamic occlusion situations. The processing modules used for object detection and tracking will be shown in detail and the performances of the algorithm discussed. The proposed approach uses an empty reference image for object extraction through image difference; the reference frame is updated continuously by a background updating module taking into account the detected objects. The tracking module is responsible for objects labeling being able to preserve objects identity even when an overlapping occurs on the image plane between different objects. A shape matching technique is used that is based on a linear Kalman filter. The system has been tested on several outdoor sequences showing dynamic occlusions among objects in order to show the validity of the approach.

This paper presents a system for automatic video surveillance applications. The system has been designed to monitor outdoor environments such as car parks or streets, providing the human operator with a symbolic description of the scene. The final task of the architecture is to automatically provide alarms when specific events of interest are detected. In this way the level of automation of the system is increased as well as overall performances. One of the main drawbacks of traditional video surveillance systems lies in the alarm generation. This task has to be visually performed by the human operators with intrinsic limitation. The possibility of having this process automated is here described within the design of an architecture capable of acquiring, processing and successfully storing data coming from one or more sensors.