Estudio comparativo de medidas de distancia para histogramas en problemas de reidentificación

Máster Universitario en Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería (SIANI)

Estudio sobre el uso de sensores RGB-D para el conteo de personas y su caracterización

[ES]Las tecnologías principales que se han utilizado son la visión por computador y los sensores de rango, es decir, las características visuales y la profundidad. Sin embargo, la aparición de sensores RGBD más asequibles, como Kinect, permite su aplicación en estos escenarios. Se aborda la utilización en entornos de interior de sensores RGBD para escenarios donde las condiciones de iluminación pueden ser variables. Se adopta una configuración cenital en el acceso a un espacio, para preservar la privacidad y facilitar la detección y seguimiento de los objetos salientes que aparecen en el escenario mediante técnicas de sustracción de fondo. Los objetos detectados son modelados, pudiendo ser descritos según las características de apariencia y geométricas como el área y volumen.

Re-identificación de personas en redes de sensores RGBD

[ES] Existe un creciente interés en detectar y analizar el comportamiento de personas en diferentes contextos empleando cámaras. Para resolver este tipo de problemas se han utilizado la visión por computador y los sensores de rango, es decir, las características visuales y la profundidad. La aparición de sensores RGBD más asequibles, por ejemplo Kinect, permite su aplicación intensiva en estos escenarios. Se aborda la utilización en entornos de interior de sensores RGBD para escenarios donde las condiciones de iluminación pueden ser variables, dificultando la aplicación de soluciones basadas exclusivamente en información visual. El uso de la información de profundidad permite ganar en robustez. Se adopta una configuración cenital para una red de sensores RGBD en el escenario de aplicación.

Resultados de los experimentos realizados de estimación de grupo de edad (2014-2015)

[ES]Resumen de resultados obtenidos en los experimentos de estimación de grupo de edad, llevados a cabo por los estudiantes matriculados en la asignatura Sistemas de Percepción del Máster Oficial de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería en el curso 2014/2015.

Reidentificación de personas y objetos mediante aprendizaje semisupervisado

[ES]La reidentificación consiste en volver a identificar a un individuo/objeto que ya se ha detectado previamente desde distintas cámaras. En este proyecto se exploran diferentes técnicas para la reidentificación de personas. Se implementan y prueban técnicas que no requieren de aprendizaje previo para realizar una ordenación inicial, al ser este tipo de métodos los que mayor aplicación tienen en un escenario real. Así mismo se usan técnicas de reordenación sobre esta ordenación inicial utilizando la información de un operador humano, aplicando entre otros métodos aprendizaje semisupervisado. Para realizar todo el proceso y facilitar la combinación y automatización de las diversas técnicas se crea un framework denominado PyReID basado en Python y OpenCV, de software libre y disponible públicamente en Github.

On the use of simple geometric descriptors provided by RGB-D sensors for re-identification

[EN]The re-identification problem has been commonly accomplished using appearance features based on salient points and color information. In this paper, we focus on the possibilities that simple geometric features obtained from depth images captured with RGB-D cameras may offer for the task, particularly working under severe illumination conditions. The results achieved for different sets of simple geometric features extracted in a top-view setup seem to provide useful descriptors for the re-identification task, which can be integrated in an ambient intelligent environment as part of a sensor network.

On the use of a low-cost thermal sensor to improve kinect people detection in a mobile robot

[EN]Detecting people is a key capability for robots that operate in populated environments. In this paper, we have adopted a hierarchical approach that combines classifiers created using supervised learning in order to identify whether a person is in the view-scope of the robot or not. Our approach makes use of vision, depth and thermal sensors mounted on top of a mobile platform.

A comparison of face and facial feature detectors based on the viola-jones general object detection framework

[EN]The human face provides useful information during interaction; therefore, any system integrating Vision- BasedHuman Computer Interaction requires fast and reliable face and facial feature detection. Different approaches have focused on this ability but only open source implementations have been extensively used by researchers. A good example is the Viola–Jones object detection framework that particularly in the context of facial processing has been frequently used.

Forecasting and visualization of wildfires in a 3D geographical information system

[EN]This paper describes a wildfi re forecasting application based on a 3D virtual environment and a fi re simulation engine. A novel open source framework is presented for the development of 3D graphics applications over large geographic areas, off ering high performance 3D visualization and powerful interaction tools for the Geographic Information Systems (GIS) community. The application includes a remote module that allows simultaneous connection of several users for monitoring a real wildfi re event.

A performance based study on gender recognition in large datasets

[EN]Gender recognition has achieved impressive results based on the face appearance in controlled datasets. Its application in the wild and large datasets is still a challenging task for researchers. In this paper, we make use of classical techniques to analyze their performance in controlled and uncontrolled condition respectively with the LFW and MORPH datasets. For both sets the benchmarking protocol follows the 5-fold cross-validation proposed by the BEFIT challenge.

A video database for facial behavior understanding

[EN]The use of new technologies in order to step up the inter- action between humans and machines is the main proof that faces are important in videos. Therefore we suggest a novel Face Video Database for development, testing and veri cation of algorithms related to face- based applications and to facial recognition applications. In addition of facial expression videos, the database includes body videos. The videos are taken by three di erent cameras, working in real time, without vary- ing illumination conditions.

Viola-Jones based detectors: How much affects the training set?

[EN]This paper presents a study on the facial feature detection performance achieved using the Viola-Jones framework. A set of classi- ers using two di erent focuses to gather the training samples is created and tested on four di erent datasets covering a wide range of possibili- ties. The results achieved should serve researchers to choose the classi er that better ts their demands.

Multi-Sensor people counting

[EN]An accurate estimation of the number of people entering / leaving a controlled area is an interesting capability for automatic surveil- lance systems. Potential applications where this technology can be ap- plied include those related to security, safety, energy saving or fraud control. In this paper we present a novel con guration of a multi-sensor system combining both visual and range data specially suited for trou- blesome scenarios such as public transportation. The approach applies probabilistic estimation lters on raw sensor data to create intermediate level hypothesis that are later fused using a certainty-based integration stage. Promising results have been obtained in several tests performed on a realistic test bed scenario under variable lightning conditions.

An study on ear detection and its applications to face detection

[EN]OpenCV includes di erent object detectors based on the Viola-Jones framework. Most of them are specialized to deal with the frontal face pattern and its inner elements: eyes, nose, and mouth. In this paper, we focus on the ear pattern detection, particularly when a head pro le or almost pro le view is present in the image. We aim at creating real-time ear detectors based on the general object detection framework provided with OpenCV. After training classi ers to detect left ears, right ears, and ears in general, the performance achieved is valid to be used to feed not only a head pose estimation system but also other applications such as those based on ear biometrics.

Competition on counter measures to 2-D facial spoofing attacks

[EN]Spoofing identities using photographs is one of the most common techniques to attack 2-D face recognition systems. There seems to exist no comparative stud- ies of di erent techniques using the same protocols and data. The motivation behind this competition is to com- pare the performance of di erent state-of-the-art algo- rithms on the same database using a unique evaluation method. Six di erent teams from universities around the world have participated in the contest.