1.1 GitHub issue || Individual poses of reference database images and view synthesis

GitHub issue link:

https://github.com/HajimeTaira/InLoc_dataset/issues/8

I am looking for individual poses of reference database images.

I went through the Inloc dataset, it looks like the pose for every reference database image has not been provided. Instead, we have a .mat 3D scan associated with every ref image, using which we can apply P3P between 2D-3D pair and obtain the desired pose as a closed form solution.

However, the pose obtained this way seems to have a slight offset.

The way I verified this is by: First using the individual 3D scan and our obtained pose (through p3p), I synthesized an image at that pose. Now since we already have our original image, I can compare the synthesized image and original image. Here are the examples:

example visualizations:

Left column is original reference image.

Right column is synthesized image at pose got from P3P (pycolmap) between 2D image and its corresponding 3D scan.

3rd column (if present): I also took a screenshot of visualizer of the 3D scan:

to know if sparsity in synthesized image is because of our view synthesis code or because of sparsity in scan. Which one is it? Probably combination of both. If there's a large hole in the synthesized image, it seems to be because of sparsity in scan. However, you can see white patches in continuous regions, this could be because of the view synthesis code.