Differential Visualizer

Description:

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 by ARCore
• *_sh.png for the implemented custom method.