Typically, training CV AI models requires labeled and diverse datasets which might contain countless elements (not only visual in nature). The collection, curation, and labeling of real-world data is inherently time-consuming, expensive, and might not be feasible. This can impede the development of CV models, slowing time to market/solution.
SDG can be used to augment existing datasets and/or bootstrap initial CV solution developments, resulting in significant time and cost reductions.
This solution leverages SD generated by previous virtual simulations, consisting of 2D and 3D assets and associated metadata (object labels and coordinates), which is then used to collect data to train a CV model.