Automatic Casting from Videos Using Deep Convolutional Neural Networks

In automatic casting applications, the aim is to accurately recognise facial regions that correspond to a same actor appearing in a movie to produce described video. In particular, this project will focus on challenging tasks of capturing and modeling the facial trajectory for each person appearing in a movie in order to predict when/where the principal actors appear. This is a challenging task because recent movies are typically high quality and faces are often occluded and their appearance varies significantly according to pose, illumination, blur, etc.

3D Heat-Map Development based on Fault Diagnosis Data

This work focuses on generation a framework to employ a set of 3D coordinates, as the input dataset to the model, and generate the 3D heat map based on the 3D shape. The generated 3D heatmap aims to define the most probable areas for fault categories on the 3D surface. To develop such a system, the 3D shape is printed and the 3D coordinates of simulated faults are recorded using a tool tracker. Then, a machine learning platform is employed to use the 3D fault datasets as the input and produce the probabilities of different fault categories on the given location.

High power all-fiber Raman laser at 1.65 ?m

Fiber lasers have become the fastest-growing laser with a projected worldwide revenue up to $1.41 billion in 2017. In particular, fiber lasers at 1.65 ?m have drawn increasing attention with potential applications in chemical sensing, LIDAR and spectroscopy. All-fiber Raman lasing technology is a promising and efficient technology to achieve high power lasing at 1.65 ?m. However, there are limited all-fiber high power sources at 1.65 ?m that are commercially available.

Simulation and Development of a new Drive system and its Control Method for Permanent Magnet Synchronous Motor (PMSM)

The proposed research project is to develop and implement a motor drive based on new semiconductor technologies such SiCMOS and GaN. This motor drive will be used in eVox bikes developed and commercialized by Procycle. The objective is to increasing the operating voltage in the system from 100V to 250V. This high voltage device will improve substantially the efficiency of the system by decreasing the losses in the electrical machine used in the bike. This improved performance will help the company to stay in head of this very competitive market.

Development of an Echo Cancellation Scheme for Full Duplex DOCSIS 3.1 Technology

Data-Over-Cable Service Interface Specifications (DOCSIS) 3.1 is a technology that utilizes orthogonal frequency-division multiplexing (OFDM) technology to provide downstream (DS) transmission capacities of up to 10 gigabits per seconds (Gbps), as well as orthogonal frequency-division multiple access (OFDMA) technology to provide upstream (US) transmission capacities of up to 1 Gbps per channel. Full duplex (FDX) communication, via the concurrent transmission of the US and DS on the same frequencies, has been proposed for boosting the US capacity.

The Characterization of Effective Dielectric Constant in Low Temperature Co-fired Ceramic (LTCC) Fabrication Process for Development of Advanced Millimeter-Wave RF Front End Systems and Components

High frequency millimeter-waves are quickly becoming essential for many industrial and commercial applications including high resolution automotive radar and high-speed wireless communication systems. While millimeter-waves offer many benefits, they also introduce new engineering challenges, due to the smaller wavelengths and dielectric properties of materials at high frequencies. One of the primary challenges relates to the accurate characterization of the substrate material parameters, which are relied heavily upon in design and simulation of millimeter-wave circuits.

Estimating Real-Estate Property Prices and Days on Market through Machine Learning

In collaboration with Western Canada Realty, we aim to develop a real-estate valuation online service for the Alberta real-estate market, especially for Edmonton. The product resembles Trulia and Zillow on a high level, which are real-estate price estimation web services in U.S. The system requires a backend that takes as input house pricing information from several sources including real-estate property assessment publicly available from the City of Edmonton website and recent house transaction records on MLS®.

Advanced design of an inverter for a new application in a telecom dc microgrid

Rapid growth of renewable energy generation and DC loads as electric vehicles and consumer electronics results in the proliferation of DC microgrids. DC microgrids are small electrical grids where energy sources and loads are connected to a main energy distribution power line using power converters.

Real-time multi-class object recognition for municipal waste collection

The use of garbage trucks with “robot arms” for municipal waste collection is becoming more and more common in Ontario and across North America. However, these “robots” have very little capability. In our collaboration with Waterloo Controls, we are working to automate the process of garbage collection. Rather than requiring the operator to acquire the truck, then steer the robot arm to pick up the bin, we are working to add automatic capability.

Intelligent surveillance for event detection

We aim to develop a video surveillance system that detects and reports events of interest to users. Traditional way is to use RGB cameras to operate in constantly and evenly illuminated areas to detect simple events. The challenge is that visual characteristics of an unconstrained scene are unstable due to illumination variations. In this project, by using RGB-D cameras, 3D position information of humans can be obtained, which is robust and insensitive to the illumination variations.

Pages