DEEP FILM Access Project (DFAP)

The DEEP FILM Access Project (DFAP) aims to unlock latent opportunities that exist within big and complex data sets generated by industrial digital film production.

Filmmaking as a process recently reached a scale and complexity where a new on-set 'data wrangling' role has emerged to manage the data generated by the camera(s) alone - further complicated by inter-relating Computer Generated Imagery and shooting in Stereoscopic 3D. Even small independent productions film on multiple cameras, increasing linearly the volume of raw material and exponentially the inter-relations between data items. Furthermore, data generated by the creative process, such as director and crew notations on quality, on the logistics and organisation of each shot, on props information and more, is recorded separately from camera data. Currently, the archival conventions for all this data contains duplication and opportunities for error; it also makes it impossible to search different kinds of data in an integrated way. As film completes its transition from photo-chemical to digital, new archival methods and processes are needed to cope with the data which also offers the potential for novel in-depth analysis of the film making process and results.

DFAP will develop an integrated process and framework for the management of all of the assets created by digital feature film production. First, we will design classifications and definitions that will standardise the description, layering and interlinking of data assets and enable them to be openly accessible online. Then, we will define a way in which this information can be integrated with the records made by all those who work on the making of the film (actors, crew, etc). Consequently, it will be possible to jointly interrogate the data generated by the cameras and the data generated by the creative process.
DFAP will explore the range of questions that this joint interrogation opens up, and how these questions might lead to film production data being used in new ways, across academic disciplines and industry professions, to challenge and expand existing knowledge and practices.

The new approach developed by DFAP will be tested and evaluated using British film director Sally Potter's latest release, Ginger & Rosa (2012), as a pilot. A period feature shot entirely on location in the UK, with a 3.5 million pound budget and a crew of over 155 members, it is almost entirely digital in all aspects. It provides an emblematic example of an industrial digital feature film production in contemporary Britain. DFAP will use the pilot to gain a sense scale and to define what its integrated system would require to be applied to other projects and materials.
The findings and outcomes of DFAP will be presented in peer reviewed journals and conferences. A project website will document the project as it develops, as well as its findings. In addition, the integrated dataset created as a result of the Ginger & Rosa pilot will be made openly accessible via Sally Potter's online archive.

The PI (Atkinson), Co-I (Evans), Mentor (Arnold) and Project Partners fully recognise the limitations on achieving substantial, long-term impact within a project under the ECR strand of the Big Data call, in which the resources (staff and non-staff costs) are capped at £100,000. For this reason, the individual pathways (described in the attachment 'Pathways to Impact') are designed to improve the chances of the research actually reaching the potential impact(s) - which are:
- a new pilot/prototype data integration model, developed from the research methodology - combining the skills of computer scientists, visual media experts, filmmakers and archive / curation specialists. The long-term aim is to develop this prototype further and apply it as a new means of searching the vast range and volume of datasets, metadata and sets of metadata associated with film production.
- enriching the data available to researchers, filmmakers, archivists / curators, scholars and those interested in films (viewers, film critics etc) by establishing a new framework of standards and methodology for recording data during the production of films - essentially to make searching for the disparate range of data on individual films much easier. This impact will be limited, initially, but the long-term aim is to encourage filmmakers to adopt a culture of thinking about the data assets as much as the visual assets from the films they create and to work with those who will be responsible for curating these assets (Museums, film production companies, universities, broadcasters and studios).
- Achieve more consistency between data generated by automated processes e.g. by filmmaking equipment such as cameras, sound equipment etc and human recorded data - film call-sheets, director and producer notes etc. This will involve the start of a process that will have a longer-term impact - using the Ginger & Rosa data assets as a case study, the PI and Co-I will be able to demonstrate the value to a variety of audiences of the importance of embedding this consistency between data produced and recorded by human intervention and automated processes.
- Encouraging academic disciplines to work together - combining skills associated with computer science with those derived from knowledge, experience and expertise in the arts and humanities. the combination of skills between the PI and Co-I will act as an example case study during the dissemination of research results activities (see Case for Support and Pathways to Impact documents - attachments) - as a means of starting to have an impact - particularly within higher education of arts-led projects that bring in advanced skills and expertise from other disciplines to improve research design, methodology and results.
- A much longer-term potential impact will be to prove the value and transferability of the research methodology and application of the research results - in particular the pilot/prototype for the data integration model (for film) - to other areas of arts and culture. For example, the principles of this research project could be equally applied to music production, performing arts such as Theatre and Opera and to other visual-interactive media e.g. gaming / video games and the serious games industry.

The core impact, running through the entire project, is the unveiling of 'embedded' or 'hidden' assets for producers and viewers of films of all types and scale - from short-films to large-scale blockbusters. Linked to this, as a longer-term aim, is to reduce the time that new filmmakers spend on pre-production, production and post-production by using the embedded data in existing films to reduce planning times, opening up skills and learning for new film producers and film crews. The impact would be reduced costs associated with filmmaking and a more enriched experience for producers, viewers of film and researchers and scholars who study film and the filmmaking process.