Deep learning-based application for construction and mining industries exploiting drone data

Drone technologies become more robust, and easier to use across important civil sectors, such as construction and mining, with the assurance to deliver unparalleled performance and consistency in every operation. However, one of the significant challenges in integrating drones into civil applications is related to data; i.e. lack of standards for high-quality data collection, complexity […]

Read More
Automating query-focused text summarization

The partner organization Croesus aims to provide its customers with a software solution for providing a natural language explanation of fluctuations in investment returns. The proposed research project offers an artificial intelligence-based solution which would provide such an explanation by summarizing financial documents in a way that answers the user’s question. An example of such […]

Read More
Classification des profils d’attaquants et attribution de l’attaque

La plupart des travaux actuels sur la détection des intrusions se concentre sur la détection et l’analyse des attaques. Peu de travaux ont été réalisés sur l’analyse du comportement des attaquants eux-mêmes. Ce type d’analyse, appelée attribution, est important car il permet de remonter à la source de l’attaque, caractériser l’intention et répondre adéquatement à […]

Read More
Large Data Visualization in Real Time

A successful computer application can easily have millions of users, but handling the network traffic generated by millions of users is not easy. The server system has to be well designed in order to properly process large, concurrent, and distributed data streams without affecting user experience. In addition, users do not generate data constantly, they […]

Read More
Neural implicit functions for multi-image low-level computer vision on smartphones

Modern smartphone cameras commonly employ multi-image (or burst photography) for tasks related to image super-resolution and high-dynamic-range imaging. This project is focused on developing novel multi-image techniques that leverage the power of recently proposed neural implicit functions. Neural implicit functions (NIFs) are a new way to represent images not as a 2D grid of pixels […]

Read More
An AI Approach to VLT Games Modeling

Video Lottery Terminals (VLT) are common gaming machines for money usually found in various venues. These machines are anonymous and stateless and do not record any information about player’s identity. The data obtained from these machines is very complex and hard to analyse. The gaming industry is keen to understand how a game or a […]

Read More
UofT-JMIR Paper-Peer Reviewer Recommender System

Rapid and open dissemination of research is critical for the advancement of science. Preprint servers (such as MedRxiv, BioRxiv, PsyRxiv, and aRxiv) are becoming increasingly popular in health and medicine to share early research results, particularly in the context of the current COVID-19 pandemic. Given the need for rapid peer review, we need to innovate […]

Read More
Improved Deep Learning-based Sea-ice Monitoring System

This project will develop a robust software package that can be embedded with a camera system to provide an onboard sea-ice monitoring system. The software package consists of two main components: (1) Deep learning classification model, which involves a deep learning network trained to identify and classify sea-ice; (2) Lens artifact removal method, which is […]

Read More
Mapping High-Level Workflows to Low-Level Robotic Actions

We are developing a robotic platform and agent that is able to translate instructions (e.g. “bring me the newspaper”) into a series of low-level programs (moving between rooms, locating objects, interacting with objects) and executing these autonomously. Traditionally, in hierarchical reinforcement learning, tasks are executed sequentially, but we would like to focus on parallelization and […]

Read More
Research an intelligent framework for conversational AI software with the ability of financial advisors’ market insight

From simple transactions to complex investment strategies and wealth management, the customers want instant, intelligent answers to their questions, comprehensive financial knowledge education, and connection to a financial advisor on social media. To provide enhanced user experiences, we are researching an intelligent framework for conversational AI software with the ability to deliver financial advisors’ market […]

Read More
QUOREM: A data science platform for microbiology research

In the genomics era, microbiologists are collecting data at an accelerating pace. We have created a data science platform called QUOREM that significantly reduces the technical burden experienced by microbial ecologists associated with managing ever-increasing quantities of data. We believe that public and private research groups in Canada could benefit our platform, and the ability […]

Read More