zhang-snowy
commented
1 year ago I found a dimension error after modifying the data set configuration code and path, so I would like to ask if you can send the test code of T&T to my email address? Thank you very much and look forward to your reply
Hello, Dr. @JiayuYANG Thanks for your amazing work and releasing your code. I wonder the experimental setting about self-sup. on Tanks and Temples training set. The Table 4 in your paper indicated the model self-sup. trained on T&T training set achieved an excellent performance compared with the model on DTU dataset. However, in your paper, there are no details about the experimental setting about the self-sup. on T&T training set.
The captured images of T&T benchmark contain more background regions compared with DTU dataset. But in the training stage, the model should focus on the foreground region, I guess. Is there any data pre-processing when self-sup. on T&T training set.
Looking forward to your reply. Many thanks!
Below is one of the images of Ignatius and its ground-truth point cloud 👇