Question about 'finetune'

[hito-Chen]

Dear author, thank you for your excellent algorithm. I have tested the examples with the data set you supposed. However, the result may not be as good as the paper shown. In city_car_test, the car is a little deformed and looks abrupt on the road. Is there anything I missed? Another question: How does the 'finetune' work？When I add 1 group of images and add 1400 groups of images in 'unpaired.txt' to finetune, the result is the same. Some results are attached. Looking forward to your reply, thank you! Best regards