Despite advances during the last decade, the recognition of deformable objects seen from different viewpoints remains largely an open problem for computer vision. Aspects of the recognition problem include deciding whether an object category, such as a face or a car, is contained in an image, identifying its location, outlining the image region it occupies and estimating its pose and the pose of its parts. There are few methods that can address all of these tasks, while the problem of performing them simultaneously for hundreds, or thousands of categories,.as humans do, has not been addressed in the literature so far. Our goal in this project is to introduce a hierarchical and probabilistic approach to high-level vision by developing appropriate object and image representations. Our objectives are versatility, namely building models able to deal with several visual tasks and scalability, that is introducing an approach that is extensible to large-scale recognition. We propose to use Hierarchical Compositional Representations (HCRs) that can account for the hierarchical nature of visual objects by modeling them in a recursive manner: structures at each level of the hierarchy are obtained by a probabilistic composition of structures at the level below, until, at the lowest-level of the hierarchy, modeling the image. A successful development of HCRs can address both versatility and scalability: Due to their hierarchical nature HCRs are able to cope with a broad range of vision problems, as their lower levels reach out to the image information -and can thus perform segmentation- while their higher levels can be constructed at a level of abstraction that allows the modeling of whole object categories, instead of simple object instances -so as to perform recognition. At the same time, as HCRs model objects recursively, parts at a certain level can be shared among multiple objects. Part sharing can then result in detection algorithms whose complexity is sub-linear in the number of objects used. We thereby aspire to develop an approach analogous to the one that led to the development of practical, large-scale systems in the problem of speech recognition: extracting a generic low-level signal representation, finding a small set of common mid-level parts and learning to combine them together into high-level structures in a probabilistic manner. We will address all aspects of the problem, including the development of appropriate mid- level representations, learning hierarchical models and detecting objects in images. We will focus on techniques that guarantee the efficiency and scalability of our system, most notably string-based mid-level representations that will be exploited both during inference and learning, efficient inference algorithms relying on combinatorial optimization, and machine learning techniques that can deal with hierarchical representations. We thereby aspire to develop a system that will efficiently recognize multiple object categories simultaneously, while requiring a small number of training images.

The grant proposal CGOSD (Corporate Governance, Organization and Sustainable Development) is set on two types of premises: - first, the widely shared assumption that in the face of the recent failure of the Copenhagen summit and of the multiple crisis – social, economic, financial, ecological – signalling the limitations and risks of contemporary management systems, building more balanced forms of resource management from a social and ecological perspective is necessary; - second, a more original consideration and a core foundation of the CGOSD program, that combining social, economic and ecological objectives in sustainable development policies is far from trivial and actually raises significant paradoxes and difficulties. Accordingly, CGOSD is established with the aims of using the analytical frameworks and tools of management sciences to identify, characterize and theorize the forms of resolution of ‘sustainable development’ paradoxes that organizational actors might build through managerial practices in the context of organizations characterized by varied governance systems. Our field inquiries are indeed geared towards diverse forms of organizations including small enterprises (part 1), large multinational corporations (part 2) and alternative non-profits and NGOs (part 3), on the basis of the hypothesis that the goals and decision systems embedded in the governance of these three types of organizations are likely to influence the ways in which organizational actors solve the paradoxes of sustainable development policies. CGOSD is organized along a matrix structure: ten tasks are vertically distributed across the three main parts of the program and simultaneously interconnected by nine horizontal comparative themes. By adopting an approach based on paradoxes – from the Greek para (against) and doxa (opinion) – we follow a critical management study perspective running against the consensual view of sustainable development held by the current managerial wisdom, that tends to focus on ecological issues and thus, overlook the complexities of balancing economic, social and environmental considerations. Such approach is channelled through a neo-institutionalist perspective on managerial practices and contextualized with reference to the Global Value Chain framework. CGOSD draws on a broad set of competencies held by 31 researchers of varied status inside the Coordinator Lab ERFI (University of Montpellier 1), and in the Partner Labs LGI (Ecole Centrale de Paris) and ICI (University of Brest) respectively contributing 11 and 6 researchers. Such partnership provides the program with a strong knowledge base on small firms, large corporations and non-profits in relation to sustainable development policy issues and the expertise of AgroParisTech researchers affiliated to ERFI in Montpellier. The program will benefit from the international research networks built by the three laboratories. Part 1 on small firms is established on the basis of a partnership with the Management School of the University of Ottawa, and its extension is envisioned under a comparative European analysis with six partner universities in Europe. Part 2 on large firms builds on close relationships with the Cultural Political Economy Research Centre of the University of Lancaster and the Otto Suhr Institute - Center for Labor Relations of the Free University in Berlin. Part 3 on non-profits and NGOs will draw on a number of partnerships established through the AgroParisTech members of ERFI with the Brasilian research Institute INPA (Instituto Nacional de Pesquisas da Amazonia), the World Wide Fund for Nature, and the International Union for Conservation of Nature, among others.

The objectives of all development efforts in computational mechanics of solids and fluids for 20 years by the international community, aim at improving predictions of numerical simulations to better reflect the physical reality and to make more robust predictions with respect to uncertainties and to the complexity of the systems modeled. This research is part of the challenge of developing new methodologies and advanced tools for stochastic modeling, for the quantification and for the propagation of uncertainties in computational mechanics of solids and fluids for a single scale or for a multi-scale modeling. Today, some of these issues are addressed and some less, and there are many results, formulations and methodologies available. However, these available techniques essentially apply to low dimensional stochastic modeling and they cannot easily be extended to the high stochastic dimension case. Moreover, there is no constructive method today to address stochastic modeling and its identification in high dimension. This project addresses key questions, unresolved to this day, regarded by the international community as challenging problems which must be solved, for a single or for two coupled scales of stochastic modeling, and which are: (1) The parametric stochastic modeling in high dimension of mechanical system parameters in solids and fluids mechanics, their quantification and their identification by solving inverse stochastic boundary value problems. (2) The nonparametric stochastic modeling of model errors related to the construction of a reduced-order model of the computational mechanical model in presence of a parametric stochastic modeling in high dimension of the computational model parameters, their quantification and their intrinsic identification performed by solving inverse stochastic boundary value problems. (3) Methods and formulations to analyze the propagation of high dimensional parametric stochastic models through linear and nonlinear computational models in fluid and solid mechanics, with or without considering a nonparametric modeling of model errors. (4) Methodologies and formulations for stochastic multi-scale modeling corresponding to the coupling of two scales with stochastic models in high dimension for linear and nonlinear solid mechanics. The application fields covered in this project are representative of the major challenges facing today in modeling and multi-scale simulation of complex systems in computational fluid and solid mechanics. The methods developed in this project will allow the handling of problems that are currently regarded as extremely difficult or even unreachable. The tools and methods developed in this project could be applicable to the numerical simulations conducted in the industry to optimize the design and performance (1) for automative vehicles of the future, (2) for aerospace and space technologies, (3) for complex systems of energy production, particularly concerning the seismic risk assessment on nuclear power plant, (4) for composites with complex microstructures, including living tissues, (5) for microsystems and even nanosystems, (6) for fluid dynamics, for solid mechanics and also for fluid-solid coupling system where multi-scale and multi-physics modeling are most often required. The proposed project represents an important gap in the knowledge and methodologies in this domain and is not a simple incremental contributions from previous contributions. It corresponds to a true technological breakthrough with respect to the capabilities of current methods to address numerical simulations of complex mechanical systems in presence of uncertainties and random media. The project explores new directions and methodologies related to unresolved issues of high dimensional stochastic modeling in computational mechanics.

Given the significant development of the urban areas in the last decades, recent parliamentary reports pointed out the existence of important territorial disparities in the organization of care by underscoring “deficits and defects in the health system”. The “Hospital, Patients, Health, Territories” Law of the 21st July 2009 launches a reform aiming at ensuring an equal access to the health care system from any point of the territory and at ensuring a more effective care delivery. The SAMU system (Emergency Medical Aid Service) created 30 years ago in order to organize the urgent medical aid at the level of each department, has to adapt itself to this reform. In France, the handling of calls arriving to the SAMU system involves a medical diagnosis. The medical decision, in the most serious cases, consists in sending a SMUR (Reanimation and Emergency Mobile Service) team on scene. However, since its creation, the SAMU-SMUR system does not have appropriate performance indicators, its performance being evaluated only at the volume of activities realised by the system. In urban areas, where the problems of accessibility are important, the analysis of the current system identified some deficiencies: the ratio of SMUR teams arriving on scene in less than 10 minutes after the reception of the call was lower than 21% (observation from SAMU 94 data). The SAMU system is therefore a perfect example of urban service illustrating the gap between the hospital and the sustainable city concepts. For this system, since urban constraints are very strong, it is difficult to guarantee an access to the care within the period of time recommended by the medical review of literature (direct relationship between mortality and time). The objective of this project is to optimise the organisation of the service provided by the SAMU-SMUR system by using a systemic approach taking into account all the elements of this complex care delivery system in an urban context. We therefore aim at working on the whole organization and the medical strategy of the SAMU system by deploying a multi-disciplinary approach based on the most relevant and recent scientific methods and technologies in order to respect the target times for critical pathological situations. The method will be developed by using data from pilot departments having maximum urban environmental constraints. It can then be deployed to the other French urban areas. The project consortium consists in: the SAMU 94 and 4 research laboratories (EA 4390 of UPEC,, LGI lab of ECP, COGIT Lab of IGN and LVMT of ENPC.).. The scientific program is structured over 7 tasks: 1 - definition of the objectives 2 - identification of urgent medical situations by new technologies 3 - quantitative performance evaluation of the global system 4 - definition of the system architecture and data flows 5 - simulation of various scenarios to optimize the static structure of the system 6 - simulation of dynamic scenarios in order to propose real time options to physicians 7 – development of organizational recommendations, a computer aided tool that enables to redefine the organization and a demonstrator for real-time operations The 24 months duration project will provide 6 deliverables: deliverable 1: methodological guidelines for the optimization of SAMU operations deliverable 2: an optimization model enabling to improve SAMU operations in urban areas deliverable 3: recommendations for a better identification of medical emergency situations deliverable 4: specifications for a decision-making software to be used by the regulating SAMU physician deliverable 5: design of a demonstrator illustrating the information flow, the interfaces between systems, their update and the tracing of data deliverable 6: recommendations towards the town planning process in terms of accessibility needs for medical emergency services