Using OCI, Twelve Labs trains video models to understand videos like humans do
The AI company uses Oracle Cloud Infrastructure (OCI) for the massive computing resources needed to train video foundation models on hours of video.
“By training on Oracle Cloud Infrastructure, I would say we gained 5X to 10X training efficiencies, which allowed us to go to market at a faster pace than we had expected.”
Twelve Labs is creating artificial intelligence models that can understand videos the same way humans do. Its models are used by developers and enterprises to search massive video datasets to find specific moments, such as searching an entire basketball season to find a particular player dunking, or sifting through thousands of hours of surveillance footage to locate a break-in. The unique challenge of training multimodal AI models is that they learn by watching many hours of video content, and thus require an incredible amount of compute resources. Twelve Labs trains the models using Oracle Cloud Infrastructure AI Infrastructure, with performant GPU node configurations and CPU memory storage, clustered using a low-latency RDMA network. Using OCI to train AI models, Twelve Labs gained 5X to 10X efficiencies, allowing the company to go to market sooner than expected.