Format d'image | EPFL Graph Search

Au cinéma et en vidéo, le format d'image est un abus de langage courant qui désigne en fait le « rapport de forme » de l'image ou de l'écran ; ce facteur désigne les proportions du rectangle d'affichage, parfois appelé « rapport de cadre » (« ratio » en anglais, ratio image, ratio écran, ratio projection) d'après l'anglais « aspect ratio » : c'est le rapport entre la largeur et la hauteur d'un photogramme ou d'un écran. Par convention, ce rapport largeur/hauteur s'écrit « largeur »:« hauteur ». Ces données peuvent provenir de mesures ou, dans le cas de la vidéo ou de l'informatique, du nombre de pixels pour les deux dimensions, c'est-à-dire la définition d'écran. Au cinéma, les formats d'image encore actuellement utilisés pour la projection argentique sont majoritairement 1,85:1, et 2,39:1, connu sous le terme de « format large anamorphosé x2 ». Les principaux formats utilisés avant eux ont été :

le format natif du cinéma muet (1,33:1), aussi appelé 4/3 à la télévision ou en DVD ;

Electronic transport properties of vertical and lateral graphene-based heterostructures

Roberto Urcuyo Solorzano

Graphene has some drawbacks, like the absence of an electronic band gap which leads to only small gate switching ratios, and its fast carrier recombination that limits its use in optoelectronics. In order to overcome these issues, the aim of this thesis was to tune the properties of graphene by two different modification approaches. The first approach involved chemical functionalization to create lateral graphene heterostructures, whereas the second approach was directed toward vertical heterostructures that combine graphene with another 2D material. For the former case, two novel chemical functionalization methods were investigated. The first method relied upon hyperthermal molecular ion collisions with 4,4¿-azobis(pyridine). This enabled the covalent functionalization of graphene with a functionalization degree of up to 3%, as demonstrated by Raman in combination with XPS. AFM studies revealed that the functionalized sheets retain their topographic integrity. Thus obtained stripe-like patterns of covalently graphene within extended graphene sheets enabled enhanced on/off ratios upon gate switching. In the second functionalization, OsO4 was used to selectively introduce hydroxyl groups to graphene. Best results were obtained using UV light activation, which yielded graphene of high covalent functionalization degree, as concluded from its optical transparency and gate-induced on/off switching ratios of up to 500. Temperature dependent electrical measurements revealed 2D hopping as the dominant transport mechanism. The first type of vertical heterostructure, gr-TiOx-Ti diodes were fabricated and their electrical properties in the dark and under visible light illumination studied. In contrast to conventional MIM diodes, the performance of the graphene-based diodes was found to increase with decreasing thickness of the oxide insulator. Bias-induced modulation of the work function of graphene was identified as the key to the operation mechanism of the graphene diodes, enabling them to reach a very high asymmetry and nonlinearity (9000 and 8). The diodes compete well with state-of-the-art MIM diodes. Furthermore, the graphene-TiOx-Ti diodes could be operated as photovoltaic cell, with a maximum open-circuit voltage of 0.3 V and a short-circuit photocurrent of 14 nA under global illuminationt. This finding constitutes the first proof-of-principle of hot carrier extraction from graphene, based on the photovoltaic effect. The second type of vertical heterostructures was based on black phosphorus (BP). As a first step, thin BP sheets were combined with n-GaAs into novel pn devices. They displayed pronounced rectification behavior that, in the low bias regime, approaches that of ideal diodes. Moreover, when operated as photodiodes they reached external quantum efficiencies (EQE) above 30% under higher reverse bias. On this basis, the bP was then combined with gr into a p+/p heterojunction. Thus obtained devices achieved a maximum internal quantum efficiency (IQE) and responsivity of 13% and 10.7 mA/W, respectively. This IQE is the highest thus far reported for diodes comprising BP and another 2D material. Moreover, with the aid of a top gate it was possible to modulate both, the photo-conversion efficiency and photocurrent generation mechanism of the black phosphorous-graphene diodes. Together with the possibility to tune the photocurrent response by the thickness of the black phosphorous, these diodes emerge as promising photodetectors.

EPFL2016

Numerical study of transitions in Taylor-Couette flow

This thesis is divided into two parts. In the first one the creation of a complete numerical tool, from the mesh generation to the data treatment, is presented. However time consuming this part has been, this is not the most important one. The second and more important part is concerned with the description and physical interpretation of the results obtained with the numerical tool previously developed. The numerical tool, which exists in its quasi-final version since mid-1997, fully satisfies the following requirements. It is specifically designed to study the flow in the annular space between coaxial, differentially-rotating cylinders of finite length. The spectral element method is used for the space discretization. The study of transition requires the high accuracy warranted by this type of method. The numerical scheme has also to be efficient. This requirement is satisfied, due to the fully-explicit time scheme adopted, where both diffusive and non-linear terms of the Navier-Stokes equations are treated explicitly, and the direct inversion of the pseudo-Laplacian matrix applied to the pressure. This inversion is performed in the most efficient way with a fast diagonalization technique. In the Reynolds numbers range we are interested in, the time-step limitation due to the linear viscous term is only slightly more stringent than the one due to the non-linear term. The last requirement that has been fulfilled has been to design as simple a code as possible. The entire code is constructed from a number of well-known algorithms, fitted together to enhance efficiency. However, the way we regularize the boundary conditions is new. It represents the physics more precisely than Tavener et al. [65]. There is a second original feature in our code, which is linked to the time scheme. We derived our time discretization from the scheme of Gavrilakis et al. [31]. Their scheme cannot be applied to cylindrical coordinates as it is. We thus had to modify it. The second part of the thesis consists of the numerical study of the first transitions of the Taylor-Couette flow in a finite-length geometry. The aspect ratio between the length of the cylinders and the gap between them has been chosen equal to twelve; this is small enough for the effects of the upper and lower boundaries of the flow to be significant. It is believed that these end-effects play a non-negligible role in the transition of the flow. On the other hand, the aspect ratio is large enough to make qualitative comparisons with the infinite-length case, which has been studied extensively both theoretically and numerically. The transition process depends on a relatively large number of parameters. Our investigation focuses on the case where the cylinders rotate in opposite directions. The study of counter-rotating Taylor-Couette flow for a large but finite aspect ratio is the main originality of this thesis. Furthermore, the physical mechanism of the appearance of the second bifurcation in classical Taylor-Couette flow, where only the inner cylinder is rotating, is described in the light of the non-linear interaction between velocity and vorticity.

EPFL1999

Person Retrieval in Surveillance Videos Via Deep Attribute Mining and Reasoning

Peixia Li, Zhiye Wang, Zhen Wei

Person retrieval largely relies on the appearance features of pedestrians. This task is rather more difficult in surveillance videos due to the limitations of extracting robust appearance features brought by the cross-view and cross-camera data with lower image resolution, motion blur, occlusion and other kinds of image degradation. To build up a more reliable person retrieval system, recent works introduced appearance attribute models to describe and distinguish different persons with high-level semantic concepts. Despite the progress of previous works, the value of utilizing appearance attributes is still under-explored. On one hand, existing methods lack for concise and precise attribute representations that are specific for each attribute category and, in the meantime, are able to filter noisy information in irrelevant spatial locations and useless patterns. On the other hand, correlation and reasoning between different attributes are neglected, which could generate more useful information and add more robustness to the retrieval system. In this paper, we propose an Attribute Mining and Reasoning (AMR) framework which is capable to handle the issues in question. The AMR makes better use of appearance attributes with two main components. First, the AMR disentangles the representations of different attributes by localizing their spatial positions and identifying their effective patterns in a weakly supervised manner. To achieve more reliable localization, we propose the Attribute Localization Ensemble (ALE) module that is consisted of multiple localization heads and a voting mechanism. Second, we introduce the Attribute Reasoning (AR) module to correlate different attributes together with the global appearance features and discover their latent relations to generate more comprehensive descriptions of pedestrians. Extensive experiments on DukeMTMC-ReID and Market-1501 datasets demonstrate the effectiveness of the proposed AMR framework as well as its superiority over the existing state-of-the-art methods. The AMR model also shows great generalization ability on the unseen CUHK03 dataset when it is only trained on Market-1501 dataset.

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC2021