Synthetic Data Generation for Training Object Detection Models
, Senior Solutions Architect, NVIDIA
, Solutions Architect , NVIDIA
Highly Rated
"How much data is enough?" is a common question when fine-tuning or training your own object detection models. In cases where data collection is a limiting factor, we can utilize synthetic generated data. Omniverse Replicator makes generating this 3D data streamlined in a single application, with the ability to modify the appearance and format of our data. This lab will use a food manufacturing example to prove the power of Omniverse in generating synthetic data to train your model. You'll start by getting hands-on experience with writing/perfecting a Replicator script within Omniverse. Then with your output data you will use a short PyTorch script to fine-tune a pre-trained model. We'll finish by testing how well our data does in training this model. This lab highlights one of the ways in which deep learning tools and Omniverse can be utilized together for streamlined deep learning workloads.
Prerequisite(s):
Basic computer vision knowledge, some Omniverse experience would be helpful but not required Internet bandwidth sufficient to support the Omniverse client/server stream (performance will vary)
We highly recommend the latest version of Chrome, a mouse, and a second monitor for optimal experience.
Please disregard any reference to "Event Code" for access to training materials. "Event Codes" are only valid during the original live session. Explore more training options offered by the NVIDIA Deep Learning Institute (DLI). Choose from an extensive catalog of self-paced, online courses or instructor-led virtual workshops to help you develop key skills in AI, HPC, graphics & simulation, and more.