MultiChoice, University of Pretoria drive AI skills

Sibahle Malinga
By Sibahle Malinga, ITWeb senior news journalist.
Johannesburg, 10 May 2024
University of Pretoria students have gained expertise in audio and video processing using artificial intelligence.
University of Pretoria students have gained expertise in audio and video processing using artificial intelligence.

The partnership between MultiChoice South Africa and the University of Pretoria (UP) has helped students to develop artificial intelligence (AI) and machine learning (ML) skills.

Established in 2018, the collaboration between the higher learning institution and the video entertainment group sought to apply AI and ML at various points throughout MultiChoice’s value chain.

This includes content creation, understanding what content to offer customers and improving interactions with MultiChoice customers across different platforms.

MultiChoice says the university has over the last few years been providing the company with human capital development, research and development, consultancy services and managerial support. In return, MultiChoice has been offering industry guidance, financial support and employment opportunities to promising students, ensuring a symbiotic relationship.

Through projects, UP students have gained expertise in audio and video processing, recommender systems and automated translation, among others.

“The automatic genre labelling of movie trailers using video and audio content involves training a machine learning model on a dataset of movie trailers, where each trailer is labelled with its corresponding genre (action, comedy, drama),” says MultiChoice.

“The model learns to identify patterns and features in the visual and auditory components of the trailers that are indicative of specific genres. Once trained, the model can then be used to automatically classify new, unseen movie trailers into their respective genres without human intervention.”

Spearheaded by graduate student Kyle Pretorius, the project demonstrates the practical application of ML in streamlining content selection processes for platforms like DStv. It automates thumbnail selection through genetic programming and convolutional neural networks, allowing for significant time and cost savings.

Participating students range from final year and Honours students, to those studying Master’s and PhD degrees in engineering, data science and computer science.

Joseph Mervitz, an engineering Master’s Degree holder and beneficiary of the partnership between the UP and MultiChoice, explains: “I thoroughly enjoy working on projects at MultiChoice. The wide technical landscape provides for a variety of different projects and challenges. I enjoy the supportive environment that allows for growth and development, and contributing to an exciting and vibrant company culture.”

According to MultiChoice, the partnership also paved the way for novel research avenues in the Department of Electrical, Electronic and Computer Engineering at UP.

Through topics spanning automated analysis of audio-visual content, to forecasting weather-related impacts on service delivery, the partnership remains at the forefront of driving technological innovation in the broadcasting industry, it says.

“Research outcomes stemming from this collaboration have not only enhanced the academic pursuits of sponsored students, but also hold the skills and insights of chair holders − professors Nelishia Pillay and Pieter de Villiers − providing influence of industry-academia partnerships in shaping scholars.”