search

xxlong0 / SparseNeuS

SparseNeuS: Fast Generalizable Neural Surface Reconstruction from Sparse views
MIT License
319 stars 16 forks source link

About Traing data Prepare #10

Open hastaluegoph opened 1 year ago

hastaluegoph commented 1 year ago

Hi, xiaoxiao, I just want to train the model from scratch. I download the mvs_training data and DTU data which you mentioned in the data part.And then I run the train code, I find the resolution of the depth is different from which you use here. depth The depth which loads from pfm format is 160*128, so I guess that the depth should be preprocessed here. Do you have plan to publish some detailed info for prepare dataset? Or how to use own data to train the model.

flamehaze1115 commented 1 year ago

The GT depth is not needed for training. We load the gt depth of DTU, just to check whether the network predicts the correct geometry in training. You can just remove the depth loading part in the dataloader for convenience.

Aiedails commented 1 year ago

Would you like to treat this as a bug? See that if all depth part is some kind of "debug" part, then it's no need for the release code to contain depth computation, or it's easy to turn off. On the other side, if you want paper readers better understand how you experiment, maybe it's more proper to publish the same dataset you use and mention it in your paper in my opinion. It's hard for a reader(me) to remove all parts containing depth because they seem to be everywhere :(

guangxuwang commented 11 months ago

bad code！