Xuan Ju

Hi, @qinb! Thank you for your focus. We plan to release the training code of SmoothNet [here](https://github.com/cure-lab?type=source) this month. After this, we will add the training code into MMHuman3D. Please...

Sorry for forgetting to check the ignore_list.txt and thank you for your answering! I list some other videos as follows: - gBR_sBM_c05_d04_mBR0_ch06  - gBR_sBM_c04_d06_mBR2_ch05  - gBR_sBM_c03_d06_mBR4_ch08 

Hi, @jinfagang ! Thank you for your focus! We didn't see such a mismatch in our code environment. Would you provide us with more detail for your question?

Hi, @theneuronprogrammer ! Thank you for your focus! Do you mean how to denoise the predicted poses without sampling? If so, simply changing the sample interval to 1 would work....

Hi, @loboere ! Thank you for your focus. The strength used in our paper is to control the unmasked region, thus seems not suitable for your requirements. I think you...

Hi, @AGI0102 ! Thank you for your focus. As mentioned in [this issue](https://github.com/TencentARC/BrushNet/issues/10), we use a segmentation mask-based ckpt in our hugging face demo. Thus the generated object shape should...

Hi, have you check whether you close the nsfw checker?

Can you try changing a VAE? I also encountered this problem in training SDXL, and changing the VAE encoder works for me. I don't know whether it would also happen...

Hi, @Arcxml ! Thank you for your focus. The hugging face space uses the segmentation mask checkpoint, thus it will force a boundary in the mask region. I recommend you...

I think the prompt will influence the results. But as I said, I believe the most important reason is the ckpt used in hugging face space. You can try locally...