hi，i want to ask some question

[wudizuixiaosa]

First of all, I'm glad to see your amazing works,，Can your model apply to the conversion of the style of underwater images to that of land? If I want to do so, what do I need to prepare? For example, I now have a ready-made underwater fish data set

[JarrentWu1031]

Thanks for your appreciation. I wonder if you mean style transfer or image-to-image translation. For the former, you could just download the pre-trained model and do the testing. For the latter, you may have to use some kind of image-to-image translation model and add CCPL during training.

[wudizuixiaosa]

Thank you for your timely reply. For example, cyclegan can complete the conversion from day to night, so I think CCPL can also complete the conversion from underwater to land. I wonder if what I said is clear? I have seen your general transformation from actual images to artistic styles, which is not what I said. So I want to ask if I want to complete the situation like I said, should I retrain?

[heraclitusj]

You're welcome. Here's the thing. In this paper, we propose CCPL and SCTNet. SCTNet is designed specially for style transfer, which means the broader image-to-image translation tasks might go beyond its power. However, the generic loss (CCPL) we proposed in the paper, is proved to be useful in image-to-image tasks. For example, we show horse-to-zebra results when adding CCPL on top of CUT model. You could also add CCPL on CycleGAN during training if you're interested. 

…

------------------ 原始邮件 ------------------ 发件人: ***@***.***>; 发送时间: 2022年8月1日(星期一) 下午4:04 收件人: ***@***.***>; 抄送: ***@***.***>; 主题: Re: [JarrentWu1031/CCPL] hi，i want to ask some question (Issue #4) Thank you for your timely reply. For example, cyclegan can complete the conversion from day to night, so I think CCPL can also complete the conversion from underwater to land. I wonder if what I said is clear? I have seen your general transformation from actual images to artistic styles, which is not what I said. So I want to ask if I want to complete the situation like I said, should I retrain? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[wudizuixiaosa]

Thank you very much! I will try it soon