Project AI-79: Water Level Depth Logger
Project Client: Opus Petroleum Engineering Ltd.
Project Description: Current techniques to monitor well water level use pressure transducers in the wellbore and leave them hanging there which endangers the well equipment. To address this issue, Water level Depth Logger is a non-intrusive and sound-based sensor that can be placed on the well cap. The data is stored on cloud as well as locally on SD card. The users are easily able access the data to view historical well level data. Additionally, the device has the feature that can detect when a pump is activated and track the rate of level change while pumping.
Project CG-22: Using IoT Devices (Raspberry PI) To Count People For Occupancy or Traffic
Project Client: UBC Cloud Innovation Centre
Project Description: Have you ever found yourself wandering around a busy public area, searching for a quiet study space or a place to eat, only to realize everything is already taken? In a world as connected as ours, why is it still so hard to know how crowded these places are before we get there?
We have implemented an IoT solution employing an edge device fitted with a camera, transmitting images to our deep learning model for comprehensive analysis. The resulting occupancy data is stored within a database and seamlessly presented on a user-friendly web application powered by React. This setup empowers users to effortlessly monitor real-time occupancy statistics with just a few clicks, facilitating timely decision-making.
Our solution not only displays current occupancy but also provides historical occupancy trends of the past 3 months for users, allows administrators to select specific sub-areas to monitor occupancy and view cost.
Project CG-39: Detecting Buried Metallic or Non-metallic Anomalies
Project Client: Rampart Detection Systems Ltd
Project Description: Our project is a Universal Buried Anomaly Detector. Regular mine detectors rely on metal detection but many mines are non-metallic. Our project overcomes this challenge by analyzing the electric field to find buried mines of all compositions, big and small. This project aims to be both cost-effective and simple to master for the untrained individual.
Project CG-78: Precise Detection of Free Street Parking Using AI and Video Processing
Project Client: UBC Digital Multimedia Lab, Dr. Panos Nasiopoulos
Project Description: Finding parking in major cities is often time consuming and difficult. Drivers are frequently forced to circle around city blocks in search of available street parking, which unnecessarily wastes time, contributes to traffic congestion, and burns fuel which emits more greenhouse gases. Our team aims to tackle this problem in collaboration with the UBC Digital Multimedia Lab, who focuses on the research and development of deep learning for diverse real-world applications.
We created an end-to-end prototype system that detects available street parking and shows drivers the locations of available spaces on a mobile application. This involved training a powerful YOLOv8 deep learning model to detect available street parking from video streams taken by in-car cameras. This parking availability data, along with location data, is then verified by robust decision-making algorithms hosted on Jetson Nano edge devices placed at intersections. Once validated, the data is aggregated in a cloud database. Any driver can then use our front-end mobile app to easily access up-to-date locations of nearby available parking spots within a certain radius of their location.
Project JM-46: Bass Guitar Pluck Type Classification Using a 2-D Piezo Pickup
Project Client: Yamaha Guitar Group Inc
Project Description: Our team worked with Yamaha to develop a new iteration of a piezo pickup system an electric bass. The main goal of this project is to contribute to the music industry by investigating the benefits of adding a second dimension to the standard piezo pickup design, and answering the question of whether or not they are worth the investment.
One major challenge was adapting the 2D piezo mount to withstand the tension of a string, and making sure the mount can be easily fit on the bass’ bridge. Another main aspect of the project’s design was developing both a 1D and 2D machine learning algorithm that is able to discern between different types of pluck, to see how the extra dimension affects the model’s accuracy.
Project PN-14: Open Data: Aerial Drone Footage- Lost Hiker Challenge
Project Client: UBC Cloud Innovation Centre
Project Description: Our product is dedicated to speed up the process search and rescue teams identify traces of lost personnel’s from voluminous drone footage. Currently, they are viewed by human or with the aid of basic outlier detections algorithms such as color analysis, we propose a solution that seeks to automate this process, to be faster and to be more accurate.
Our products utilizes states of the art deep learning architecture trained one of the most comprehensive publicly available dataset. We also provided the option to upload a video directly to our cloud server for ease of use.
Project SF-07: Interactive Augmented Reality Factory
Project Client: UBC MANU Program
Project Description: The Interactive Augmented Reality Factory project was proposed by Dr. Christoph Sielmann, an Assistant Professor of Teaching at The University of British Columbia. Being a representative of the Manufacturing Engineering Department, he identified that a challenge for manufacturing engineering students was conceptualizing the space required for solutions that they learn about in their program. The Interactive Augmented Reality Factory empowers manufacturing engineering students by providing them with the opportunity to interact with novel environments in limited lab space through configurable Augmented Reality (AR) simulations.
Our team created an AR platform using Unity that supports both Android and iOS devices, ensuring cross-compatibility and accessibility. Keeping the limited software training of the teaching staff in mind, we designed a simple simulation configuration framework that uses JSON configuration files uploaded through a web browser to generate simulations.
A key feature of our application is the inclusion of 10 dynamic objects, specifically designed to allow for a variety of simulations. These models include metal ingots, conveyors, buttons, indicators and more. Using our configuration process, teachers can easily allow students to view different types of simulations using their personal devices, enabling them to gain a deeper understanding of the manufacturing vision.