Predict normal difference for each pixel point.
First, download and process the data as described in ../data, and link it to here by ln -s <PATH-TO-DATASET> data. Then, go through the regression.ipynb.
We provide not only normal difference comparison but also per-line classification (on whether structural) result comparison between ground truth (LSD(Pytlsd)) vs predicted for each image of the dataset. Go through the visualization.ipynb.