This is a simple script used for evaluation of a master thesis.
The script simply searches through the source folder, which is set at the beginning of the code, for .png
images.
The goal is to compare ground truth images with images generated by googles ARCore
algorithm and images generated
from the implemented master thesis method, which relies on spherical harmonics.
This script takes the ground truth images, creates heatmaps of differences to the images produces by the two algorithms
and calculates MSE
and
PSNR
which is stored as a json file into the source folder.
Note that the script ignores black pixels when calculating the errors. The script was created for renderings of objects with a black background. We only want to compare the foreground!
Images must follow this naming for the script to work:
*_gt.png
for ground truth images*_ar.png
for images generated byARCore
*_sh.png
for the implemented custom method.