As we all know, data is the backbone of modern artificial intelligence and machine learning applications. The accuracy and performance of these models heavily rely on the quality and diversity of the data they are trained on. However, traditional datasets often suffer from inherent biases, originating from various sources, such as demographics, socio-economic factors, and geographical locations. These biases can have far-reaching consequences, leading to unequal representation and skewed decision-making processes.
Synthetic Data Generator is a program which generates a synthetic video of a given input, being able to change race, gender, or age by using Next3D. It is also capable of relighting the output, frame by frame, by utilizing RI_render_DPR.By using this program, we can easily eliminate geographical biases due to a limiting dataset.
reenact.2.mp4
The results of relighting can be seen in relighting-results