In this note, I’ll present an unsupervised algorithm that can extract three-dimensional features from an ordinary two-dimensional image, and detect edges within the image, thereby extracting two-dimensional shape information, in each case in polynomial time.

A research note explaining the algorithm, together with the code, is available on my researchgate homepage here, and I’ve also attached a PDF.

The relevant scripts are also attached as PDFs below. Any missing scripts can be found in previous posts, and in the “Algorithms” tab.

extract_3D_features

identify_fine_fast

maximize_std_dev_fast

test_measures_approx