An attempt to enable Pony, and now IL, to generate more accurate helmets.
Trained on photos of various figures in tinted motorcycle helmets, male and female. This means the model isn't good for visible faces.
Description
Trained on a wider dataset.
-Removed most bikes from the background of training images in an attempt to reduce the model's inclination to generate a vehicle.
-Pruned images with a duck-bill style lip on the back of the helmet.
-Tinted visor shouldn't be needed as a trigger word anymore, since it's assumed, but it most likely won't negatively affect results if used.