Stephen Jolly
Research Interests
Stephen Jolly has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Graeme Phillipson
BBC Research & Development, United Kingdom
2 shared publications - Michael Evans
BBC Research and Development, Salford, United Kingdom
2 shared publications - Craig Wright
BBC Research & Development, United Kingdom
1 shared publication - Em Shotton
BBC Research & Development, United Kingdom
1 shared publication - Jack Allnutt
BBC Research & Development, United Kingdom
1 shared publication - Lianne Kerlin
BBC, London, UK
1 shared publication
Latest Publications
- Open Access
- 12 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 77-78
AI in Production: Video Analysis and Machine Learning for Expanded Live Events Coverage
- 12 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 77-78
Open AccessAI in Production: Video Analysis and Machine Learning for Expanded Live Events Coverage
- Open Access
- 12 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 76
Old School: An 8K Multicamera Shoot to Create a Dataset for Computational Cinematography
In our paper, we describe the rationale for, and the process and outcomes of, a shoot that BBC R&D carried out in 2021 to create the Old School dataset, to further research into AI automated editing systems. We introduce our way of thinking about these systems, and explain why we believe a widely available dataset is necessary if research into these systems is to reach its potential. We then explain why a new shoot was necessary to generate such material. We proceed to give an overview of the creative process, explaining aspects of the shoot that differed from conventional TV production in order to produce a dataset with the desired technical and editorial characteristics. We describe the planning and pre-production, the shoot itself, and post-shoot work to process the material and generate metadata. We finish with a summary of the material and metadata that have been included in the asset bundle for the dataset, and an overview of how we decided what to include in this bundle. This paper was published as a BBC R&D White Paper in July 2022 [3].
- 12 June 2023
- Published by Association for Computing Machinery (ACM)
- p. 76
Open AccessOld School: An 8K Multicamera Shoot to Create a Dataset for Computational Cinematography
In our paper, we describe the rationale for, and the process and outcomes of, a shoot that BBC R&D carried out in 2021 to create the Old School dataset, to further research into AI automated editing systems. We introduce our way of thinking about these systems, and explain why we believe a widely available dataset is necessary if research into these systems is to reach its potential. We then explain why a new shoot was necessary to generate such material. We proceed to give an overview of the creative process, explaining aspects of the shoot that differed from conventional TV production in order to produce a dataset with the desired technical and editorial characteristics. We describe the planning and pre-production, the shoot itself, and post-shoot work to process the material and generate metadata. We finish with a summary of the material and metadata that have been included in the asset bundle for the dataset, and an overview of how we decided what to include in this bundle. This paper was published as a BBC R&D White Paper in July 2022 [3].