Department of Computer Science

Capstone Project's 2020 Landing Page



Product Owner:

Codehesion

Traffic Light Optimization With Machine Learning
By Team Aegis

Traffic in big cities can quickly become an issue and in today's times having people reset traffic timers is just not cutting it anymore. The idea involves creating a traffic optimisation system using machine learning, image processing and tracking. The basic concept would be all traffic light intersections would have a set of cameras each aimed towards a specific road, these cameras would detect the number of cars waiting at the traffic light and feed that information into a bigger system. With collection of these the system can start making smart decisions in terms of traffic light durations. If it sees a large number of cars coming from a specific direction, it could increase the lights for that direction to prevent congestion. With a large enough set of these this system should be able to increase the flow of cars and decrease traffic based on the current time of day and level of traffic. A virtual based system would suffice as a proof of concept. In this instance actual hardware like cameras, raspberry pies and servers wouldn’t be needed. The system could be developed and demoed in something like Unity with a virtual traffic system.



Product Owner:

epiuse

Artifact
By Team 5Bits

The Fake News Detector is a tool to be used by both News Consumers and News Distributors to act against the spread of fake news. Machine Learning will be used to detect fake news and make users aware of its presence. We will also maintain databases of known fake news distributors as well as the common fake facts that they spread.



Product Owner:

Agile Bridge

Anaroc
By Team Code-Sum-Moar

Anoroc traces the sources, routes and transmission points of communicable diseases by using spatial data analytics, mobile apps, artificial Intelligence and community participation.



Product Owner:

EpiUse Labs

Courier Driver Tracker
By Team Ctrl+Alt+Elite

This system is the tracking of a courier fleet, to monitor and recognizing the driving behaviors. This system should monitor courier drivers and be able to detect patterns and suspicious (outside of the normal) behaviors.



Product Owner:

EpiUse

Object ID
By Team Ctrl+Intelligence

In a unique landscape in respect of crime prevention we are trying to utilize technology – and by extension “smart” technology to help in the proactive monitoring and reaction to potential crime incidents. As part of the current technology stack we utilize number plate recognition, making use of either proprietary camera based (edge) recognition routines and/or cloud routines. Our requirement is to utilize imagery (stream and still images) to build a dataset with which to constitute an electronic signature of each vehicle – allowing us to distinguish using unique features (scratches, dents, stickers etc) rather than just a numberplate.



Product Owner:

EpiUse Labs

Data Visualization
By Team Doofenshmirtz Evil Inc

Use an Interactive Genetic Algorithm (IGA) to suggest visualizations for dashboards and drill-downs. Suggestions should be based on the data as is or preprocessed through other Artificial Intelligence (AI) algorithms.



Product Owner:

EpiUse Labs

Truckinit
By Team Enigma

In short, we’re aiming to design an Uber-like web/mobile app for the world of logistics that’ll connect Truck Fleet Managers and their truck drivers with potential delivery jobs and clients.



Product Owner:

EpiUse Labs

Plushie Designer
By Team Fellowship of the Git

A plushie is a soft toy with an outer fabric sewn from a textile and stuffed with flexible material. For this project you will create a website that allows users to create and share their plushie creations.



Product Owner: Quant Solutions

Quant Solutions

3D Model Binary Vision
By Team Flap Jacks

3d model from binary vision camera with web based 3d colour rendering site. The intention is to scan a physical dental model with stereo vision, in parts into a system, that builds allows rotation and viewing.



Product Owner:

EpiUse

Gym Moves
By Team Lockdown Squad

Most people dislike going to an over-crowded gym during peak hours or upon arrival realising that their spinning class has been canceled. The idea is to build an app that tracks real-time traffic in a gym, so that users can make informed decisions about when to visit the gym.



Product Owner:

Evo Academy

Swift
By Team Lumiqon

Swift aims to create a convenient, flexible, and scalable interface suitable for both customers and restaurants. Customers will have the opportunity to order, track and pay for their food immediately. The experience will be akin to ordering food from an application such as Uber Eats, with the contrast that it will be an in-restaurant experience. The application will also use Augmented Reality to show real live photogrammetry models of the food, making the menu come to life. Swift will aim to accelerate the speed of execution and simplify the food ordering process of restaurants resulting in minimum delays and confusion.



Product Owner:

AWS

Watchdog
By Team Lynk Solutions

Security in South Africa is for ever changing. With many neighbourhoods having adopted large walls and electric fences, the need to control access and notify owners or security of breaches. In this project we will create a proof of concept by using facial recognition to create an authentication system. This system should able to alert the owner and security company on any breach or potential breach that has occurred or is likely to occur. Providing the owner/security with information such as a video or image of the intrusion.



Product Owner:

Inside Data

Unidel
By Team Memory Inject Lamas

A unified delivery application with a built-in fleet management system. This application is used to track and control delivery of goods for multiple service providers, using mobile devices that scan QR codes, take photographs and track GPS location while including real-time route delivery optimisation and fleet management. This application is for the driver delivering the items, the courier company and the courier head office, which includes the fleet manager and call center as well as the end customer who will receive the couriered parcel(s).



Product Owner:

EpiUse Labs

The Amazing App
By Team NewTech

There are many types of location-based games where the gameplay involves players being at a certain location and allowing them to progress and perform actions based on the location they are in. Examples of location-based games include Ingress, Pokemon Go, Wizards Unite and CluedUpp games. For this project, you will be tasked to create a mobile application to run a location-based game and a web application that can be used to create the location-based game. Examples of location-based games for this project include an Amazing Race event or a CluedUpp detective game. You will create markers on a map in which players have to physically visit in order to solve riddles and retrieve information.



Product Owner:

EpiUse Labs

Alfa
By Team Pyraspace

For project ALFA you will create an artificial intelligence-powered automated log file analyser. After training, ALFA should be able to scan a given log file and then intelligently suggest solutions to the problems encountered.



Product Owner:

Fabi

Fabi
By Team Runtime Terrors

TPCP database analytics and interface for data capturing, monitoring and modelling of forest pests and diseases.



Product Owner:

EpiUse Labs

Argus
By Team Sigma

Home security is an increasingly important concern in our country. For this project you will be creating a security system that will provide alerts based on intelligent pattern recognition.



Product Owner:

AWS

Scubamate
By Team AV

Divers around the world rely on a little book containing information about their dives, to some divers this is valuable to keep around as it records information about dive that you can go back and reference, such as dive site conditions, people you dove with, how many dives you have completed (Which becomes important when you want to start a professional dive journey), as such divers log entries are often signed by an Instructor (Course specific dives must be), a Dive Master or just a dive buddy.



Product Owner:

5DT

Lightbot
By Team Gradient

Traffic congestion has a negative impact on productivity – whether through delayed deliveries or delayed personnel. Traffic lights are used to control the flow of traffic at intersections but can have unfavourable effects if the control policy is inefficient. LightBot is a proposed system that will adaptively control traffic lights in an attempt to maximise the flow of traffic.



Product Owner:

EpiUse Labs

Puzzle Generator
By Team Prometheus

There are many types of wooden puzzles. Some require a sequence of steps to open a box. Some require you to assemble or construct different shapes. For this project you will be creating a web application that can create and simulate such puzzles.



Product Owner:

EpiUse

Project Tree
By Team B-Team

The project entails the development of a project management system. The system represents a project visually as a tree structure, with each “leaf” being a project task, and each branch being a task dependency. The core idea of the system is to allow for all task dependencies to be captured accurately and clearly, to ensure effective project management.



Product Owner:

EpiUse

STAT
By Team Visionary

By using the latest innovations in AI, IOT and mobile platforms, we want to make tracking your productive time on a daily basis as easy as possible. By removing half the effort of keeping track of what you’re doing, when you’re doing it and for how long, it leaves you with more time to focus on improving performance in other areas, and free you up for more fun things.



Product Owner:

EpiUse

Animal Track Recognition
By Team Zenith

ERP Rangers require an application to assist with tracking animals via their tracks left on the ground or their droppings in the field. Users should be able to identify animal tracks / droppings by taking a photo of the object in question and running it through a cloud endpoint capable of identifying the species of animal it belongs to. The application should assign a certain probability to images matching with an animal, if uncertain it should provide users with additional matching options, ranked by probability. Once matched, the software should display facts about the animal along with other tracking signs. This should include animal prints, droppings, pictures, possible habitat and common behaviours. A GPS log can be created for the user attaching geotags to the images of tracks taken, this should include date and time with the location. Feasibility analysis and testing will take place at one of our covered reserves. This data could be used to determine if other rangers have seen animals in a similar area and display a heatmap to determine possible areas where the animal may currently be.



Product Owner:

Iminsys

Traffic Light Optimization
By Team Alpha

The aim of this project is to develop a virtual traffic network which can model almost any real world problem in simulation. The main target is to be able to optimize the traffic flow through a well calculated timing of traffic light changing. This would be archived by the use of AI operations working on time series data.