Deadline: Friday 26th April
Length of internship: 3 months
The Data Science Institute is pleased to announce that, as part of their partnership with the BBC, we are able to offer another exciting internship opportunity. The projects is titled, 'Quality evaluation of AI-produced video' and will take place between June and September 2019 with a stipend of £1500/month.
Applicants must have permission to work in the UK, and be able to take an interruption to their studies (if applicable) in order to complete the internship.
To apply, please email email@example.com with a two page CV and a short (<500 word) statement explaining how their skills match the proposed research project, by 11.59PM on Friday 29th March 2019.
To help us understand how AI/ML technologies could transform the business of producing media, the AIP team are developing a prototype system to automatically edit live panel shows and similar performances. We are working to enhance the quality of the system’s rendered output; more attractive framing, more natural editing, showing the right sequence of things that viewers most want to see. However, other than anecdotal, subjective impressions, how can we know that we’ve done something to improve the output, and how will we know when output quality is ‘good enough’ for certain content, and certain use cases? For a panel show, for example, we might find that despite spending a lot of effort trying to optimise framing of faces within the image, that's hugely less important than making sure there's a comfortable frequency of wide, establishing shots, or not cutting too quickly.
Research/Activity to be undertaken
The graduate on this project will work closely with both the engineers and research scientists on the project. The work will involve a practical review of previous work on factors affecting the quality of edited video sequences, and any ways in which it has been assessed in the past. In addition, experimental work with real viewers and footage should allow us to understand how well these methods and factors are indicative of perceived quality for the sort of video material produced by AI.
- Review previous work on the psychovisual impact of edited video sequences to build an understanding of the relative importance of various factors
- Help devise and conduct initial empirical user research to inform computational models of the subjective quality of a machine-edited video sequence
- Experience of designing studies/experiments in Human Computer Interaction to analyse human action using qualitative and quantitative responses
- Literature review
- Working with participants in behavioural/psychophysical user research
Expected Outputs and Deliverables
- Expert written literature review
- Completed experimental study and initial findings
Additional Skills/knowledge brought to us by hosting this intern
- Deeper psychological understanding of factors affecting perceived quality
- Building repeatable experimental methods for assessing progressive improvement in algorithmic performance in AI production/ video processing
In order to ensure that your application is processed in time, please send your CV and short statement to firstname.lastname@example.org by 11.59PM on Friday 26th April.