Majed El Helou

@tonyzzzt Check this [comment](https://github.com/NVIDIA/apex/issues/83#issuecomment-439123560). Quick fix; you can upgrade to PyTorch >1

For the h5f you can check the code from our other project here [https://github.com/IVRL/w2s/blob/master/code/generate_h5f.ipynb](https://github.com/IVRL/w2s/blob/master/code/generate_h5f.ipynb) The code below should help you out for the data preparation, of course you first need...

We originally published the code necessary for the SFM experiments only and did not publish all our analysis code. We received many requests for the PSD calculation so we added...

In general, we noticed empirically that applying SFM on 50% of the training data, while the rest is augmented by only noise maps, provides better generalization strength to unseen noise...

> "the coefficient of epsilon is not sqrt-ed": It is in the original repo: [this line](https://github.com/andreas128/RePaint/blob/8e2a0552a0cc306e623da807d353915f1af17726/guided_diffusion/gaussian_diffusion.py#L368) But you are right the coefficient square root is missing from the paper itself.

> How about the off by one timestep part tho, is it intended or maybe a small negligence ? Yes I also spotted some off by one errors, sorry for...

Hello Glad you find the dataset useful! 1. The same FoV was consistently used across all captures. We are keeping most metadata private because the objective is to enable single-image...

Hi yunchao, only the CVPR 2021 data contain depth maps, not the ECCV 2020. Each image (`.png`) file has an associated depth (`.npy`) file with the same name. Hope this...

Hi, We understand there can be limitations with GDrive, and as we are receiving considerable interest for the dataset we are now working on setting it up on a permanent...

Hi hanwei, We have set up this mirror: [https://datasets.epfl.ch/vidit/VIDIT_train.zip](https://datasets.epfl.ch/vidit/VIDIT_train.zip) Best regards, and good luck for the competition if you are participating!