Manning cameras with operators is often too expensive for productions and as a result static shots are chosen to increase the number of perspectives of a stage. Whenever moving objects are involved static shots are clearly inferior to dynamic ones since objects are never at the optimal framing point. To resolve this problem, the Automatic Control Laboratory of the ETHZ in Cooperation with ZHdK Cast / Audiovisual Media and dreicast have developed automated robotic solution (robotic camera) for video production companies based on state-of-the-art neural networks, predictive model-based control and machine learning.
The intelligent camera system can track moving objects without the need for a hands-on camera operator. Therefore, the producer can execute repeatable tasks over long periods of time whilst following precise, reproducible movement trajectories. Moreover, the producer can take dynamic shots of the stage without the need for a dedicated operator that adjusts framing. This increases quality without increasing cost.
The RoboDOP comprises of a robotic head and a processing unit that runs vision intelligent software to control the orientation and settings of any attached camera and lens. The movement of the camera is designed to imitate the way human operators frame moving objects. Following the set up process –attaching the RoboDOP head to the tripod and connecting it to the live feed of the camera and a network control cable to the control computer provided by Seervision – the RoboDOP calibrates to the camera and lens, ready to receive commands from the system operator.
The product is applicable in live communication events like conferences, corporate presentations, webcasts and workshops. In future, the product is aimed to expand its offering to allow for controlling multiple follow-camera use cases for events like Sports and artistic events such as music concerts, theatre plays where close up dynamic shots are a fundamental part of the video storytelling.
Nikos Kariotoglou, ETH Zurich
Reto Hofmann, ETH Zurich
Conrad von Grebel, dreicast GmbH
Martin Zimper, Cast / Audiovisual Media, ZHdK
Eric Andreae, Cast / Audiovisual Media, ZHdK
January to December 2017
Research and Development Project CTI
Gallery 1 (Level 4)