Système de reconnaissance d’objets pour drone reposant sur un modèle d’apprentissage profond

L’objectif de ce projet pilote est d’équiper des drones longue portée avec un système complémentaire de navigation assistée par l’intelligence artificielle. Plus précisément, ce système permettra aux drones de mieux percevoir leur environnement et d’y réagir plus intelligemment. La principale préoccupation concernant les drones volant hors visibilité directe (HVD) est reliée au partage de l’espace […]

Read More
Développement d’un algorithme de classification des nuages de points lidar aéroporté par apprentissage profond pour traiter des classes « non sol »

La compagnie XEOS Imagerie œuvre dans le domaine de l’acquisition de données lidar (Light Detection and Ranging). Elle désire extraire automatiquement les points associés au sol et aux objets à partir du nuage de points 3D brut de l’acquisition lidar. Deux stages précédents ont permis d’identifier les points « Sol » et « Eau ». […]

Read More
Development of a Privacy-preserving Infection Risk Assessment Protocol

In light of the recent COVID-19 pandemic, contact tracing has been found an effective measure to mitigate the spread of the corresponding SARS-CoV-2 virus. The aim of contact tracing is to identify, notify and potentially quarantine persons who have been in close contact with infected individuals, thereby breaking the chain of infections. Next to manual […]

Read More
A decentralized peer-to-peer lending platform using blockchain-based distributed ledger technologies

A Peer-to-Peer (P2P) lending platform allows small businesses to obtain working capital funds through an online system, where individuals can anonymously pool together funds for investment. Instead of relying on a third-party website for crowdfunding (such as Kickstarter and Indiegogo), we propose a decentralized platform for P2P. Unlike traditional P2P lending which is usually done […]

Read More
Unsupervised Machine Learning Approaches to Define the Underlying Structures in Data

In many real-world machine learning problems, inadequate labeled data and extensive unlabeled data are usually available. Normally, unlabeled data is collected routinely, though the high cost associated with labeling the data is high for generating the labeled training dataset for supervised learning. Moreover, the lack of domain experts and the time-consuming data labeling, especially in […]

Read More
Autonomous Surveillance and Inspection using Drone for Oil and Gas Pipelines

Using drones and sensors to conduct autonomous monitoring has shown benefits in many applications, including urban infrastructure planning and maintenance. In this research, we focus on a less studied problem of detecting hazardous and illegal events along oil and gas pipelines. We aim to deliver a real-time Internet of Drones (IoD) system, which integrates the […]

Read More
Data synthesis using generative adversarial network

This project is about synthesizing data using generative adversarial network (GAN). Unlike conventional studies which use anonymization techniques for removing private information of individuals, we use variants of GAN architectures for crafting new records contextually similar to real records in the legitimate dataset. We plan to run exploratory experiments on public datasets to provide enough […]

Read More
Generating Contextually Appropriate Followup Questions

The goal of this internship is to design and build a model which can generate contextually appropriate questions during a market research survey. This project will build on existing work, and will extend it by incorporating commonsense reasoning, long-range memory, and possibly other features The model will be tested with real humans using an online […]

Read More
5G-DetNet/TSN for Industrial Automation

The 4th industrial revolution, Industry 4.0, promises to create completely flexible production lines with automated-guided mobile robots. To enable this transformation, 1/0 devices (e.g., sensors and actuators) require not only a 5G Ultra-Reliable Low-Latency Communication (URLLC) but also a deterministic and bounded low latency with a guaranteed data delivery and extremely low data loss. In […]

Read More
Reinforcement Learning Algorithm for Traffic Signal Control to Reduce GHG Emissions

Persistent traffic congestions which are varying in volume and duration in the cities are not just the source of frustration for drivers, but also one of the biggest sources of greenhouse gas emissions in the world. Those types of congestions cannot be adequately resolved by the traditional traffic signal controllers. Breeze Labs Inc. in cooperation […]

Read More
Ammara Syeda

Due to the fact that the IT Department at the City of Saskatoon is facing multiple challenges due to COVID-19, the current service catalog and ticketing systems are lacking efficient self-service options, which are vital during this time. Using my previous UX knowledge, I hope to assess the current systems in place and collaborate with […]

Read More