opengvlab / ddps Goto Github PK

Hi! Thanks a lot for releasing the codes :)

Do I understand correctly that the main model is in https://github.com/OpenGVLab/DDPS/blob/main/mmseg_custom/models/decode_heads/segformer_head_unet_fc_head_single_step.py#L128 ?

Am I correct that first the discrete class labels are embedded into real-valued vectors, put into the unet which outputs real-valued logits? (and that corruption is done in the discrete label space)

Thanks for your excellent work! In Algorithm 1, the input of the denoiser network contains mask, h, and time. But in code, the input of the unet network only contains mask and time. Could you please tell me which one is the right version? Thanks a lot!

I can't find the segformer_mit-b2_512x512_160k_ade20k_20220620_114047-64e4feca.pth in segformer_b2_ade20k_singlestep.py? Can you provide the full download address? thanks！！！

Hi, thanks for the great work!
I have a quick question, could you tell me the difference between the mmseg and mmseg_custom directory in this repo?
If I want to use my own dataset, in which directory should I choose?
Thanks again!

Hi @Zeqiang-Lai,

I find your work very interesting, but struggle to infer from the paper which priors you exactly used to include the different geometric constraints. Could you please give some more detail/references about that?

Thanks in advance!

Please release your code

Hi, I am trying to do the inference but seems the debug_q_pred module is missing. Could u provide them please, or assist me in how to get them?
Thanks!