Closed maybeLx closed 1 year ago
Hi @maybeLx , that's a good question. If I remember correctly, to draw Figure 1a, I computed roughly the scales after stacking several CDSConv layers. So, Figure 1a does not show the estimated scales using only one CDSConv layer. During the experiments, I tried options for choosing the scale candidates for each CDSConv layer. So, I think Figure 1a was drawn after applying 2 CDSConv layers in line 211, 212, but the scale candidates might not be the same.
Please read the last paragraph of Section 3.3. We explained that "To have a low computational complexity, the number of candidate scales in CDSConv layer should be small, we only use 2 or 3 candidates. However, the searching scale-space is expanded profoundly when multiple CDSConv layers are stacked in CDSFNet".
I hope this will answer your concern. Please let me know if you have more questions!
Thanks, I understand
Dear author, It is really excellent work in MVS. But during reading your paper, I have some misunderstanding about your paper Figure1, fig (a) shows many scale selections for pixels in different resolution images, however you only have 3 5 7 11 kernel size to choose, I have no idea about the scale 17.