levihsu / OOTDiffusion

Official implementation of OOTDiffusion: Outfitting Fusion based Latent Diffusion for Controllable Virtual Try-on
Other
5.53k stars 809 forks source link

Is realily you need random crop garment latent ? #120

Closed Gzzgz closed 7 months ago

Gzzgz commented 7 months ago

the paper this part:

图片

图片

In the inference time,the code not drop garment latent when uncond.

levihsu commented 7 months ago

Please read more carefully. Dropout is performed for the input of the outfitting UNet, i.e., garm_latents. Thanks

Gzzgz commented 7 months ago

OK, this is here? 图片