DTU(3 views)

[wujinhu1999]

Regarding the training of DTU with 3 viewing angles mentioned in your paper, I would like to ask if you finally used all the pictures in scan65, that is, 49 pictures for training?

[niujinshuchong]

Hi, we use 3 views for the DTU 3 views experiments. We use all the images for the dense views case.

[wujinhu1999]

So for Figure 1 in the paper, how many pictures in scan65 did you use for training when you were training the three views of the DTU dataset? Which three views did you choose?Looking forward to your reply.

[wujinhu1999]

Can your work realize that only 4 pictures are input for training, such as two side views, one front view and one top view, and then get better reconstruction results?

[niujinshuchong]

Hi, I think it might also work. You could simply try it out.