#loss-weights

Question 1: Single choice

Neural style transfer optimizes the pixels of a synthesized image to minimize content_weight * content_loss + style_weight * style_loss + tv_weight * tv_loss. Your result keeps too little of the content image's layout and is covered in high-frequency speckle. Which change targets both problems?

Raise content_weight and raise tv_weight1
Lower content_weight and raise tv_weight2
Raise content_weight and lower tv_weight3
Raise style_weight and lower tv_weight4

Timed test