Result seems quite bad when i train in my own dataset.

[justinday123]

Hi, thank you for your nice work. I trained the model for "sparse scribble adapter", but result seems bad. What might be the reason?

first gif is ground truth and second gif is generated result.