yiqun-wang
yiqun-wang
Hi! sorry for the late reply, please check our updated paper https://arxiv.org/abs/2206.07850. We explain this in detail. Thanks
I just provided pre-trained models for better inference. For custom data, not sure if it is caused by the hardware environment, moreover it is hard to judge the impact of...
Hi, we provide the link of the dataset in the run.sh
Hi, thanks for your reminder. I have updated the datasets in the link.
Hi, I am sharing the obj mesh model of the NeRF synthetic dataset that we used in the provided link. https://drive.google.com/drive/folders/1hgbsUmtkjKbw8exLXF-DtcRWoVp6GQ_Y?usp=share_link Thanks
Hi, we don't provide the config with masks, but you can just modify _mask_weight_ in the config and provide the mask files for training. Best wishes.
Thank you for your attention. For this model, you can set the scale of the position encoding of the base implicit function to be smaller during training.
Hi, this issue should be that the surface has not been well reconstructed at the beginning
Hi, I just want to make the wavelet normalized, and the sum of the normalized wavelet is 1.
Hi, you can check another repo https://github.com/yiqun-wang/HFS to obtain the ground truth mesh