Nvidia has filed a patent for a system that automatically generates ground truth data for training and retraining machine learning models used in object detection. This system uses a combination of machine learning and object tracking algorithms to create new ground truth labels, reducing the need for expensive and time-consuming human labeling. This data can also be used to train other models and increase the robustness of training data sets.
