Kinect | EPFL Graph Search

Kinect, initialement connu sous le nom de code Project Natal, est un périphérique destiné au matériel Microsoft avec la Xbox 360 pour la V1 et la Xbox One et Windows depuis la V2 permettant de contrôler une interface sans utiliser de manette. Il a été conçu par Microsoft en . Le mot-valise « Kinect » est issu des mots anglais « kinetic » (qu'on peut traduire par « cinétique ») et « connect » (qu'on peut traduire par « connecter »). La sortie européenne a eu lieu le . Le , deux mois après sa sortie, Microsoft annonce avoir vendu 8 millions de Kinect, dont un million en seulement 10 jours. Le , Kinect entre au Livre Guinness des records comme étant « l'accessoire high-tech le plus vendu dans un court laps de temps » avec 10 millions d'unités vendues, soit une moyenne de 130 000 Kinect vendus chaque jour à travers le monde. Les ventes se sont ensuite érodées, jusqu'à ce que Microsoft décide de ne pas intégrer de port Kinect dans la Xbox One S sortie en

3D Shape Similarity

Fan Zhang

Microsoft Kinect, Google's Project Tango and Lytro's light field camera are all examples of 3D depth sensing reaching the consumer market. As this technology becomes more widestream, new signal processing techniques are needed to exploit this data. Recent techniques in 3D shape similarity assesment present exciting opportunities to develop new algorithms in this area. The aim of this project is to create a tutorial to these techniques using an iPython notebook. Particular emphasis will be placed on an excellent review paper [1] and book [2]. Seneca, the Roman philosopher, said “While we teach, we learn,” and creating a tutorial on these topics is an excellent exercise to understand a complex subject that is at the heart of many modern computer vision techniques. [1] S. Biasotti et al "Recent Trends, Applications, and Perspectives in 3D Shape Similarity Assessment", 2015. [2] A. Bronstein, M. Bronstein, R. Kimmel "Numerical Geometry of Non-Rigid Shapes", 2008. LCAV1556050917

2016

Wind-packing of snow: How do wind crusts form?

Christian Gabriel Sommer

Wind-packing of snow is the process responsible for the formation of wind slabs and wind crusts. These are hard layers of well-sintered snow often found in the snowpack in mountains as well as in polar regions. Wind-packing affects the local mass balance, the avalanche danger and how the snow cover interacts with the atmosphere. Yet, it is a poorly understood process. Many ideas about what wind-packing actually is have been proposed in the literature but there is almost no quantitative information about the involved physical processes available. A closed-circuit wind tunnel simulating an infinite fetch was specifically designed and built for this purpose. Experiments were performed with natural snow that was collected on a pair of wooden trays. The hardness of the snow and how it changed during experiments was measured with a SnowMicroPen. Meteorological parameters such as wind speed, air temperature, etc. were also measured. In the literature, there is conflicting information about the necessity of drifting snow for wind-packing. The wind tunnel experiments showed that drifting snow is a necessary but not sufficient condition for wind-packing. We observed no hardening during experiments without drifting snow, but not all drifting snow events lead to the formation of a wind crust. The spatial variability of the hardness of wind-influenced snow was high and this appeared to be related to the dynamics of erosion and deposition. To analyze the influence of these processes in more detail and quantitatively, a Microsoft Kinect sensor was added to the wind tunnel. This instrument measures the snow depth in the main test section with a high spatial and temporal resolution. These measurements showed that wind-packing only occurs through the deposition of snow. If drifting snow causes the snow to be eroded, there is no hardening effect on the remaining snow. Furthermore, the Kinect data allowed to explain about 50% of the observed variability of the hardness of wind-packed snow with only two parameters. Most important was the wind exposure. The snow became harder in more wind-exposed positions compared to wind-sheltered areas. We used the parameter Sx to characterize how wind-exposed a certain position is based on the upstream snow surface. A wind-sheltered position has a high Sx value and a wind-exposed position has a negative Sx value. To increase the range of Sx values, some experiments were performed with an artificial obstacle in the main test section, which created highly wind-sheltered positions. Sx alone explained almost 40% of the variability. The second parameter was the deposition rate. We found a negative correlation of -0.4 between the deposition rate and the snow hardness. Wind-packing is more efficient at creating a hard layer if the snow is deposited slowly. The effect of other parameters, such as wind speed, air humidity or initial density of the snow on the hardness was studied. However, there was either no robust trend in our data or the addition to the multilinear regression did not significantly increase the explanatory power of the model. This does not necessarily mean that these parameters are not relevant for wind-packing. The wind tunnel experiments provided quantitative information and new insights about wind-packing of snow and under what conditions it occurs. However, the question about which physical processes are most important could not yet be fully answered.

EPFL2018

Natural user interface for lighting control: Case study on desktop lighting using modular robots

Alexandre Bernardino, Stéphane Bonardi, Simon Lukas Hauser, Auke Ijspeert, Mehmet Hasan Mutlu, Massimo Vespignani

Roombots (RB) are self-reconfigurable modular robots designed to explore physical structure change by robotic reconfiguration and adaptive locomotion on structured grid environment or unstructured environment. The primary goal of RB is to create adaptive furniture. In this study, we propose a novel and user-friendly interface to control position and intensity of a mobile desk light using RB modules. In the proposed method, the user interacts with the RB with only hand/arm gestures. The user’s arm is tracked with a single Kinect having bird’s eye view. We demonstrate the effectiveness of the proposed interface in real hardware setup and discuss contributions of it.

Ieee2016

Wind-packing of snow: How do wind crusts form?

Christian Gabriel Sommer

EPFL2018