The model is trained on publicly available images of life in Urban China in 1980s.
Versions <1.0 are trained on the first training dataset consisting of only monochrome images.
Versions >1.0 are trained on the second training dataset comprising colored images