Holmes-Alan

Results 27 comments of Holmes-Alan

pose generating error when using provided silica dataset

same error here, have you resolved it?

using my own data, I can use test.py to generate 360 scene images, but I cannot extract mesh file

I know where the problem is. When training the NeRF on your own data, use downscaled images, rather than the original size. More importantly, downsampling your own images by 8...

Newbie Concerns

Firstly, you need to unzip data into the folder. Secondly, I have just upload "prepare_images.py" for testing.

is there an up-sampling layer at last?Otherwise, the spatial size doesn't change?

The input LR image is pre-upsampled by Bicubic. Please read the paper and see the details.

prepare_images

Sorry, I just uploaded it.

Unsupervised training

Unsupervision is for super-resoltuion where we only have LR images and no HR images. As for denoising, we have both noisy and clean image with same resolution provided by NTIRE2020.

i am facing this error when i try to run test.py file. would you please give me any suggestions

You need to modify the model file since we tried the network on two GPU using torch.nn.DataParallel(). If you only use one GPU, you need to change the load function...

Trouble in SMPL estimation from 3D scans

same problem here

questions about the paper

yes, we have modified this typo, and you can check the newest version for your research. Thank you

self.idle_input is not defined in modules.py

problem is fixed.