Visualization of complex configuration items Dec 2020 - Present: A configuration management solution software is a software that helps users understand and manage the configuration of complex software systems, including dependency analysis, reporting, transport, revision control and comparison. The project will aim to explore means of implementing interactive graph visualisation that will allow users to meaningfully and usefully interact with the graph. Both the visualisation from the point of view of the whole graph or a single item will be explored. The goal will be to allow users to decrease and increase the amount of detail shown based on their need by grouping items together using contextual grouping mechanisms that preserve the maximum amount of useful information while obscuring unnecessary detail. Also, the project may allow us to perform graph analysis or other machine learning techniques on the configuration.
Clothes recommendation and body shape classification Jul 2020 – Nov 2020: A new business intends to provide online personal styling advice to enable their users to make more informed and sustainable shopping decisions. It is expected that the services will contribute towards the reduction of fashion waste as people only purchase what really suits them. The project is aimed at achieving two things: 1) Developing a system that provides style, color and brand advice to users, based on the characteristics (such as body shape, budget and color features) of the users. 2) Building an approach that can be used to classify users as having one of the different body shapes. Domain: Data Science, Artificial Intelligent, Computer Vision, Deep Learning
Table detection in an unstructured document Feb 2020 – Aug 2020: A company provides a service to analyze their clients’ documents and provides them with more insight into their data. They have the vision to eliminate manual laborious and repetitive tasks; hence, they want to automatically extract specific metrics from a set of documents. Those metrics are lying sporadically on the unstructured documents, which makes them challenging to be extracted; But, most of those metrics are often appears inside the tables of the documents. Thus, a project is set to identify tables on the unstructured documents. Domain: Data Science, Graph Neural Network, Deep Learning
Topics dependencies in math Apr 2020 – Aug 2020: An online platform has developed a directed graph database of topics in math subject. Each of the topics is labelled with an associated difficulty level, and with dependencies between them determined through expert judgement. Treating these as hypothesised dependencies, the aim of the project is to find ways to collect and analyse data, generated through users’ use of the system, and to validate these dependencies between topics. Domain: Data Science, Bayesian Network, Graph Analytics
Brand recognition in a video Jan 2020 – Apr 2020: A fan-based video company provides an online platform where their users can upload videos through their mobile phone while attending a public or private event (e.g. music concert, sport match, etc.). Those videos are curated into a new film out of all the uploaded video clips in an event. To extend their business, they intend to develop a new feature that is a brand/logo detection on the uploaded clips. Domain: Artificial Intelligent, Computer Vision, Deep Learning