hht2001

> Dear authors, thanks for your interesting work and plans. However, there is one question in my mind: why you choose to use VQVAE instead of VAE? As stated both...

准确的相机位姿已经可以靠ControlNet去实现了

> > Hello, does this error occur when you want to train a model? Because this script allows you to train from scratch without a pre-trained model. Or does it...

> For what I understand it's not a controlnet but a connector used for the controlnet used in [SUPIR](https://github.com/Fanghua-Yu/SUPIR): > > > It designs an effective 620 million parameter adaptor...

> For what I understand it's not a controlnet but a connector used for the controlnet used in [SUPIR](https://github.com/Fanghua-Yu/SUPIR): > > > It designs an effective 620 million parameter adaptor...

You can find "torch.abs(pixel_prop[:,:1,:,:] - curr_lr[:,:1,:,:])" in RNN_Swin_4.py, Line 216&239. The "1" means only use the first dimension because all the dimension are same(if use the gray input). You can...

> u can process in this way: if u have a rgb image "217, 65, 43" for "r, g, b", then u can get three "gray image"-----"217, 217, 217" ,...

hi, I commit a test_video.py . u can use it to process ur own data. waiting for author's check

Thanks :)

> @KhawlahB I am trying similar things, how about your results? Relly looking forward to your response:) while compute attention in the unet, the prompt embedding will be the key...