Latent Lenses and Computational Cameras

Join us on Wednesday 2nd February 2022 at 15.00 (online) for our next research event with Daniel Chávez Heras reflecting on computer vision and photographic practice, including the example “Made by Machine: When AI met the Archive” (BBC, 2018). All welcome.

Deep learning computer vision systems are trained on large datasets of images collected from the internet. Very often, these images are photographs, which is to say they are produced through photographic processes, including lenses of different focal lengths operated at different shutter speeds. Therefore, it can be argued that this type of computer vision system aggregates and abstracts some of the epistemic and aesthetic affordances of photographic practice. In this presentation I explore the concept of a latent photographic camera in computer vision systems. Taking the BBC’s experimental television programme “Made by Machine: When AI met the Archive” (2018) as a case study, I present an experiment designed to identify the latent camera and implied optical perspective through which the BBC archive was said to be seen by machine.  


Chávez Heras, D., Blanke, T. On machine vision and photographic imagination. AI & Soc 36, 1153–1165 (2021).  

Daniel Chávez Heras is a Lecturer in Humanistic and Social Computing at King’s College London. His research focuses on the computational production and analysis of audiovisual culture, an area he approaches through a critical-technical blend of film theory, interdisciplinary design, and creative AI.
Before becoming a university lecturer, he was a research fellow at the Cultural Analytics Open Lab in Estonia, collaborated with the BBC to create the world’s first broadcast AI-TV programme, founded an international short film and digital art festival that took place entirely online, and was a digital manager for the British Council in Mexico.