Thanks for your interesting work, I tried to train the model with fern_2v.txt which works reasonably. But the model can not be trained when I set depth_loss = False and using the Nerf loss to regress RGB alone. There are only white images produced in this case. Can you inform me whether there is a proper configuration which works well for training without depth_loss, so that I can compare against the baseline methods.
Thanks for your interesting work, I tried to train the model with fern_2v.txt which works reasonably. But the model can not be trained when I set depth_loss = False and using the Nerf loss to regress RGB alone. There are only white images produced in this case. Can you inform me whether there is a proper configuration which works well for training without depth_loss, so that I can compare against the baseline methods.
Thanks