Alexis Andre

Single-view sketch based modeling

This paper presents a new sketch modeling system that is able to generate complex objects drawn from a unique viewpoint. The user draws the model in an iterative manner, adding simple parts to the existing object until completion. Each part is constructed from two construction lines (lines on the surface of the object that are planar and perpendicular to each other) whose orientation in the 3D space is uniquely determined by the system, and an optional silhouette. The system is then able to produce rough 3D reconstructions of drawings very easily by tracing over a sketch for example. Such models are perfectly suited to investigate their shade or shadow and they can be used as substitutes for more detailed models when the need for quick models is present. The user can also explore shapes directly on the system, refining the shape on the go in a oversketching way. The creation of models is very efficient, as the user models the shapes directly in the correct pose and orientation. Results show that the system is able to create complex objects without ever having to change the viewpoint.

CrossSketch: freeform surface modeling with details

This paper presents a novel technique to model a three dimensional freeform surface, in its global shape and small details, using a sketching interface, from a single point of view. In the past, most modeling systems that used sketches as input reconstructed the shape from the silhouette, and the user had few control on the inner parts of the result. In our system, we generate a grid of co-planar lines from a small number of strokes that the user drew, then we estimate the normal vector where it is constrained, and we form the surface by propagating this information to the whole grid. As a result, smaller strokes act locally to add detail, while longer strokes modify the whole surface. Our system gives a new approach to the modeling from sketches problem, and is intended to be a part of a more complex modeling system.

Peer-to-Peer Control System for DC Microgrids

We propose and implement a dc microgrid with a fully decentralized control system, using the ICT concept of network overlays and peer-to-peer (P2P) networks. Decentralization not only concerns the physical systems and control logic but also the control structure which provides the network infrastructure on which energy management is carried out. In this paper, we show how such decentralization can be achieved using P2P frameworks as underlying control structures and implemented a pure P2P to eliminate single points of failure. For this, a direct current open energy system made of the interconnection of standalone dc nanogrids is used as underlying microgrid. The power flows between nanogrids are controlled by a decentralized exchange strategy: each household can request or respond to energy deals with its neighbors without requiring system-wide knowledge or control. Using dc combined with a layered, modular software allows loose coupling which increases flexibility and dependability. The system has been implemented and tested on a full-scale platform in Okinawa including 19 inhabited houses. Real data analysis as well as simulations demonstrate improvements in self-sufficiency compared to other types of systems. Resilience against utility blackouts is proven in practice.

A Study of the Efficiency of Line Renderings of 3D Shapes

Shape perception depends on a lot of factors, for example texture or shade. When rendering a surface, the question whether the viewer correctly understands the shape of the rendered object is open. The use of different rendering methods can improve or harm shape recognition. This paper presents various styles of line drawings, three traditionally used and a new one, that we describe here, then evaluates their success at conveying the shape of the underlying object. Each line drawing is generated according to simple rules, using geometric properties of the object, for example the normal vector orientation or the principal directions of curvature. An psychophysical evaluation is then performed to compare the various styles. Participants had to adjust a probe until it matches the normal vector they perceived. Results show that the new style that we propose performs better than the principal directions of curvature, for particular cases.

Real-Time Spherical Videos from a Fast Rotating Camera

We present a new framework for the acquisition of full spherical videos using a high-speed rotating linear or area sensor equipped with a fish-eye lens. The high capture rate of the sensors allows us to reach high frame rates on the resulting videos. We describe the image processing workflows from the raw data to the full spherical frames, and we show experimental results of non-parallax panoramic videos covering the complete field of view.