Computer Vision Speaker Series

Runway is a full-stack, applied AI research company that trains and builds generative AI models for content creation. Most recently, Runway launched Gen-2, a multi-modal AI video model and first publicly available text-to-video model. In partnership with NVIDIA, Runway is working to improve their models' processing speeds and controllability, and is ensuring their tools can be used safely by creators and artists

Computer vision has become ubiquitous, accelerating human tasks and improving accuracy by giving computers the ability to “see”. In this talk, we’ll highlight the work that uses various computer vision technologies including NVIDIA Metropolis and the TAO Toolkit to identify objects of interest, track them, and build logic pipelines for analytics applications to improve warehouse operations. We’ll also talk about how we used NVIDIA Omniverse™ to create digital twin 3D assets in simulation models that can feed into the computer vision models to efficiently scale up the application.

Speakers will discuss the usage of the NVIDIA Omniverse™ Replicator in the robotics pipeline. They will showcase the usage of the NVIDIA Omniverse™ Replicator in a few projects in the agriculture, manufacturing, automotive, and space fields of robotics, as well as discuss the benefits and limitations of using synthetic data generation in robotics.

AWS HealthImaging (AHI) is a HIPAA-eligible, highly scalable, performant, and cost-effective medical imagery store. It delivers sub-second image retrieval latencies at scale powered by lossless High-Throughput JPEG 2000 (HTJ2K) encoding. The client application can retrieve medical images from AHI and decode with HTJ2K using the nvJPEG2000 developed by NVIDIA at incredibly fast speeds. Decoding on the GPU is extremely beneficial, as the image data is already in GPU memory for model training. This eliminates the need to copy data in case decoding on the CPU and enables fast HPC use cases at lower cost.