Category: Uncategorized

CeRVIM Webinar: Tirer profit des données en conception optique

Geoffroi Côté
Laboratoire de Recherche en Ingénierie Optique
Dép. de physique, de génie physique et d’optique, Université Laval

7 mai 2021, 11h00

Résumé / Abstract
En vision numérique, la modélisation de la caméra se fait habituellement au moyen d’un modèle optique approximatif, dit de premier ordre. L’analyse précise du comportement de tels systèmes optiques, de même que leur design, relève plutôt du domaine de la conception optique.

Dans cet exposé, je vais donner un bref aperçu de la conception optique, et expliquer comment le domaine peut tirer profit de méthodes axées sur les données. Je vais présenter comment l’apprentissage profond peut mener à la création d’outils concrets pour les concepteurs optiques, en combinant l’apprentissage supervisé habituel à un second signal d’entraînement basé sur le tracé de rayons optique. Finalement, je vais aborder une méthode pour intégrer la conception optique à la recherche en vision numérique.

La présentation sera donnée en français et les diapos seront en anglais.
The presentation will be given in French and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Pour obtenir le lien d’accès pour la rencontre Zoom, SVP contacter:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Parallelism for Fast and Interactive Visualization of Big 3D Data

Duc Thien Luong
Laboratoire de Vision et Systèmes Numériques
Dép. de génie électrique et de génie informatique, Université Laval

30 avril 2021, 11h00

Résumé / Abstract
In recent years, the needs for 3D visualization have significantly increased in a wide range of applications: research, industry, entertainment, transport, security, even finance and investment. All of these applications involve processing and visualizing big data. There are 3 main obstacles in visualizing the big data. First, the computation workload of the 3D data can exceed the capacity of the commercial memory core (RAM). Secondly, the popular methodology in computing requires long computation times. Thirdly, interactivity between the users and the system is not preserved when we move the displaying screen because of the system delays. The proposed parallelism method is built to overcome each of these three problems. Although we use a powerful PC to develop this method, there is no requirement for special hardware. This method only focuses on programming methodology and software solutions. In addition, we maximize the use of GPU computing and parallel programming to reduce processing time.

La présentation sera donnée en anglais et les diapos seront en anglais.
The presentation will be given in English and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Pour obtenir le lien d’accès pour la rencontre Zoom, SVP contacter:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Covariance Based Differential Geometry Segmentation Techniques for Surface Representation Using a Vector Field Framework

Mana Eskandari
Laboratoire de Vision et Systèmes Numériques
Dép. de génie électrique et de génie informatique, Université Laval

16 avril 2021, 11h00

Résumé / Abstract
In this presentation, the concepts of differential geometry traditionally applied to the segmentation of range maps is revisited in the context of implicit surface representation of unorganized point clouds. We proposed an approach to demonstrate that it is possible to combine covariance-based differential geometry and implicit surface representation methods to perform the segmentation of an unorganized point cloud (and not just a range map) into different surface types. The advantages of combining covariance-based differential geometry and implicit surface representation are that the segmentation does not require surface fitting nor does it require that all points be processed, thus reducing computational complexity. The acquisition of the point cloud data is achieved with handheld scanners used in metrology applications.

La présentation sera donnée en anglais et les diapos seront en anglais.
The presentation will be given in English and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Pour obtenir le lien d’accès pour la rencontre Zoom, SVP contacter:
Annette.Schwerdtfeger@gel.ulaval.ca

REPARTI Webinars:

A journey on your brain highways: diffusion MRI and connectomics of the future
Maxime Descoteaux
Sherbrooke Connectivity Imaging Lab (http://scil.usherbrooke.ca/)
Dép. d’informatique, Université de Sherbrooke
February 16, 2021, 12:00 – 1:00 p.m.

Understanding the world behind the image
Jean-François Lalonde
Laboratoire de Vision et Systèmes Numériques (LVSN)
Dép. de génie électrique et de génie informatique, Université Laval
March 29, 2021, 12:00 – 1:00 p.m.

Ultrasound imaging: let’s talk
Catherine Laporte
Laboratoire de traitement de l’information en santé (LATIS)
École de technologie supérieure (ÉTS)
April 22, 2021, 12:00 – 1:00 p.m.

CeRVIM Webinar: Evaluation of Skid-Steering Kinematic Models for Subarctic Environments

Dominic Baril
Norlab (Northern Robotics Laboratory)
Dép. d’informatique et de génie logiciel, Université Laval

20 novembre 2020, 11h00

Résumé / Abstract
In subarctic and arctic areas, large and heavy skid-steered robots are preferred for their robustness and ability to operate on difficult terrain. State estimation, motion control and path planning for these robots rely on accurate odometry models based on wheel velocities. However, the state-of-the-art odometry models for skid-steer mobile robots (SSMRs) have usually been tested on relatively lightweight platforms. In this paper, we focus on how these models perform when deployed on a large and heavy (590 kg) SSMR. We collected more than 2 km of data on both snow and concrete. We compare the ideal differential-drive, extended differential-drive, radius-of-curvature-based, and full linear kinematic models commonly deployed for SSMRs. Each of the models is fine-tuned by searching their optimal parameters on both snow and concrete. We then discuss the relationship between the parameters, the model tuning, and the final accuracy of the models.

La présentation sera donnée en anglais et les diapos seront en anglais.
The presentation will be given in English and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Pour obtenir le lien d’accès pour la rencontre Zoom, SVP contacter:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Radio propagation models for differential GNSS based on dense point clouds

Vladimír Kubelka
Norlab (Northern Robotics Laboratory)
Dép. d’informatique et de génie logiciel, Université Laval

13 novembre 2020, 11h00

Résumé / Abstract
Accurate geolocation of mobile equipment operating in outdoor environments is an increasingly important question in robotics and automation. Modern geolocation systems, however, rely on the crucial ability for a mobile device to receive specific radio signals at all times. As such geolocation systems are increasingly deployed in harsh or difficult environments, for example, in the presence of tall buildings or dense forests, it becomes critical to predict how the environment will impact the propagation of these radio signals. We propose a signal propagation model that can determine what areas would be favorable for global navigation satellite system (GNSS) positioning, based on a prior three‐dimensional (3D) point cloud map of the environment. The model predicts both the number of usable satellites for a GNSS receiver and the strength of the reference radio signal used in the differential GNSS scenario. We take into account both signal occlusion and absorption mechanisms, given the geometry and density of the point cloud map. The design of the model is data-driven, based on experiments performed both at the university campus and in the Montmorency forest.

La présentation sera donnée en anglais et les diapos seront en anglais.
The presentation will be given in English and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Pour obtenir le lien d’accès pour la rencontre Zoom, SVP contacter:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Design, Analysis and Preliminary Validation of a 3-DOF Rotational Inertia Generator

Jean-Félix Tremblay-Bugeaud
Laboratoire de robotique
Dép. de génie mécanique, Université Laval

15 octobre 2020, 11h00

Abstract
This presentation investigates the design of a three-degree-of-freedom rotational inertia generator using the gyroscopic effect to provide ungrounded torque feedback. It uses a rotating mass in order to influence the torques needed to move the device, creating a perceived inertia. The general working of the device is presented, along with a comparable concept using three flywheels instead of a gyroscope. Simulations are conducted to establish motor torque and velocity requirements, and the gyroscopic concept is identified as having the less demanding requirements. Preliminary experimental validations are conducted, confirming that it is possible to both reduce and increase the rendered inertia.

The presentation will be given in French and the slides will be in English.

To obtain the Zoom meeting web link, please contact:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: A Mixed Reality Interface for Handheld 3D Scanners

Jérôme Isabelle
Laboratoire LVSN
Dép. de génie électrique et de génie informatique, Université Laval

9 octobre 2020, 11h00

Résumé / Abstract
The user interface is an essential part of handheld 3D scanners. During the scanning process, it provides feedback to the user to help him operate the scanner in an efficient way. For instance, it generally displays the reconstructed 3D model in real-time to let the user know which parts of the object have been captured and which have not. Traditionally, this type of information is displayed on a 2D screen, via a graphical user interface. Instead, we propose to use a mixed reality headset. We claim that this technology is better suited for handheld 3D scanning because it allows the reconstructed 3D model to be blended into the user’s perception of the real world. To validate this claim, we developed a prototype that uses the HTC Vive Pro headset as an interface for a handheld 3D scanner based on a Primesense Carmine RGB-D Camera.

Pour obtenir le lien d’accès internet (Zoom), veuillez contacter:
To obtain the Zoom meeting web link, please contact:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Workspace enlargement and joint trajectory optimization of a (6+3)-dof 3-[R(RR-RRR)SR] kinematically redundant hybrid parallel robot

Kefei Wen
Laboratoire de robotique
Dép. de génie mécanique, Université Laval

25 septembre 2020, 11h00

Résumé / Abstract
In this presentation, the workspace and trajectory optimization of a (6+3)-dof 3-[R(RR-RRR)SR] kinematically redundant hybrid parallel robot is investigated. The inverse kinematics of the robot can be solved analytically and the singularities are easily avoidable. A workspace analysis is provided and it shows that the orientational workspace is very large. Moreover, the redundant degrees of freedom are optimized in order to further expand the workspace. An approach is developed to determine the desired redundant joint coordinates so that a performance index can be minimized approximately when the robot is following a prescribed Cartesian trajectory.

The presentation will be given in English and the slides will be in English. 

Pour obtenir le lien d’accès internet (Zoom), veuillez contacter:
To obtain the Zoom meeting web link, please contact:
Annette.Schwerdtfeger@gel.ulaval.ca

CeRVIM Webinar: Model Identification for Robotic Manipulation

Abdeslam Boularias
Robot Learning Lab
Dept. of Computer Science, Rutgers School of Arts and Sciences

20 août 2020, 11h00

Résumé / Abstract
A popular approach in robot learning is model-free reinforcement learning (RL), where a control policy is learned directly from sensory inputs by trial and error without explicitly modeling the effects of the robot’s actions on the controlled objects or system. While this approach has proved to be very effective in learning motor skills, it suffers from several drawbacks in the context of object manipulation due to the fact that types of objects and their arrangements vary significantly across different tasks. An alternative approach that may address these issues more efficiently is model-based RL. A model in RL generally refers to a transition function that maps a state and an action into a probability distribution over possible next states. In this talk, I will present my recent works on data-efficient physics-driven techniques for identifying models of manipulated objects. To perform a task in a new environment with unknown objects, a robot first identifies from sequences of images the 3D mesh models of the objects, as well as their physical properties such as their mass distributions, moments of inertia and friction coefficients. The robot then reconstructs in a physics simulation the observed scene, and predicts the motions of the objects when manipulated. The predicted motions are then used to select a sequence of actions to apply on the real objects. Simulated virtual worlds that are learned from data also offer safe environments for exploration and for learning model-free policies.

Biographie:
Abdeslam Boularias is an Assistant Professor of computer science at Rutgers, The State University of New Jersey, where he works on robot learning. Previously, he was a Project Scientist in the Robotics Institute of Carnegie Mellon University, and a Research Scientist at the Max Planck Institute for Intelligent Systems in Tübingen, where he worked with Jan Peters, in the Empirical Inference department, which was directed by Bernhard Schölkopf. From January 2006 to July 2010, he was a PhD student at Laval University under the supervision of Brahim Chaib-draa. His PhD thesis focused on reinforcement learning and planning in partially observable environments.

Pour obtenir le lien d’accès internet (Zoom), veuillez contacter:
To obtain the Zoom meeting web link, please contact:
Annette.Schwerdtfeger@gel.ulaval.ca