Our deep learning models provide us with two distinct capabilities, firstly they ensure our volumetric output is as clear and accurate as possible and secondly they allow us to capture volumetric video using far fewer cameras that conventional system.
Using fewer cameras has a number of benefits:
Using convolutional neural networks and generative adversarial networks with our groundbreaking method for training them has given us a system unmatched in its capability for capturing content in volumetric 3D.
Our system has been carefully tuned for specific types of events. This is because a general purpose algorithm which can capture any scene accurately in real-time does not yet exist. We use a process known as transfer learning to share information between models which are then specialised in individual types of capture for instance boxing or golf.
Transfer learning also informs our longer term deep learning strategy as we are confident that the models which we are building today will become the foundation of future deep learning models for far more advanced volumetric capture in the future.
Our pipeline uses GPU’s from on-site through to the cloud. Our early partnership with NVIDIA through their inception program has ensured that we have both guidance and access to hardware needed to create this kind of system.
Powerful graphics processing units like the ones that NVIDIA provides are the reason that a system for capturing volumetric content can exist today. The high performance computing capabilities that modern GPUs provide support our conventional computer vision algorithms as well as our deep learning capabilities.
5G technology promises wireless connectivity at super fast speeds. For our business the faster speeds provided by 5G networks represents a tipping point for real-time volumetric video streaming.
Our volumetric video streaming service requires a high speed internet connection at both the site where the volumetric video is being captured and any location viewing the content. 5G means that this connection no longer has to be wired. This advancement opens up a number of new possibilities for capturing and streaming volumetric content.
Our capture system, CR Capture, has been designed to be portable and quick to set up. By combining CR Capture with a 5G wireless connection our customers will be able to capture volumetric content at locations that it would have otherwise been impossible.
By pairing a 5G network connection with the next generation in augmented reality headset we expect our end users to be able to watch live events in full volumetric 3D in real-time.
Copyright Condense Reality 2019