search

levihsu / OOTDiffusion

Official implementation of OOTDiffusion: Outfitting Fusion based Latent Diffusion for Controllable Virtual Try-on
Other
5.56k stars 813 forks source link

请问大概还要多久可以释放训练代码呢，想训练自己的数据，非常感谢！ #68

Open sanbuphy opened 8 months ago

sanbuphy commented 8 months ago

如题，这是很棒的工作。 赞，向作者致敬

phanxuanphucnd commented 8 months ago

Please share the model for us to test @sanbuphy Thank you

trikim commented 8 months ago

借楼向知道的大佬请教个问题，训练的时候给模特图（gt）加噪，经过vton_unet预测噪声，MSE计算预测噪声与原噪声的loss，这个操作正确吗？因为在推理时vton_uent每次新的step，latents都是经过上次预测去噪后的latents、ori_img_latents和mask_latents这三个数据融合的，这块有点疑惑。

BaochaoZhu commented 7 months ago

@sanbuphy ，"Could you please tell me if you have the knowledge and skills to train this particular model?"

liupengcnu commented 7 months ago

借楼向知道的大佬请教个问题，训练的时候给模特图（gt）加噪，经过vton_unet预测噪声，MSE计算预测噪声与原噪声的loss，这个操作正确吗？因为在推理时vton_uent每次新的step，latents都是经过上次预测去噪后的latents、ori_img_latents和mask_latents这三个数据融合的，这块有点疑惑。

最近也在复现论文效果，可以交流一下吗？liupengucas2015@gmail.com

lyc0929 commented 7 months ago

I try to write a trainning scrpt ,u can refer this code https://github.com/lyc0929/OOTDiffusion-train.

phanxuanphucnd commented 7 months ago

I try to write a trainning scrpt ,u can refer this code https://github.com/lyc0929/OOTDiffusion-train.

Thank you @lyc0929