leonidk / direct_gmm

Project code for "Direct Fitting of Gaussian Mixture Models"

Direct Fitting of GMMs project website

This is the source code and project history for the following publication

Direct Fitting of Gaussian Mixture Models by Leonid Keselman and Martial Hebert (arXiv version here)

Overview

Almost all files used in the development and testing of this project are in this folder. The data files for the Stanford Bunny is included in bunny.

• mixture contains the modifed version of scikit-learn with the proposed techniques.
• gmm_fit.py and gmm_fit2.py contain the two sets of the bunny likelihood experiments
• registration_test.py contains the mesh registration (P2D) experiments
• Files with _extra are usually just copies for non-Stanford Bunny experiments
• gen_gmm.ipynb and gen_gmm_mine.ipynb generate GMM models from the TUM dataset, with and without uncertainty models
• reg_results.ipynb performs D2D registration between the GMM models built from the TUM dataset.