Turn abandoned patents
into your own filing
into your own filing
This invention describes a system for automatically creating training data, called ground truth, for generative AI models. It works by first analyzing input images to identify human poses and specific areas, then modifying the pixels in those areas to a uniform value. This modified image data, along with the detected poses and a text description, is fed into a neural network to produce new images that serve as 'ground truth' to train a second AI model for applications like creating AR-style full-body images of people.
Why it matters: Filed as generative AI matured, the increasing demand for high-quality, diverse training data makes automated ground truth generation more valuable. Advances in controlling generative models with precise inputs like poses and masked image areas have made this approach more effective.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI
