The Mixture of Gaussian’s or mostly mentioned as Gaussian mixture models is a probabilistic statistical technique that can be used for image segmentation. Implementations and improvements of this approach can be seen on the paper by [4] and [5].

The code present on the repo focuses on GMM and the report looks at the comparison between GMM probabilistic approach and U-Net deep learning approach.

This project was done using python (jupyter-notebook), and in the Project_GMM2020 folder there are 2 other subsequent folders and files namely:

• im - is the folder that has the visualization gif and other images, used in the repo and jupyter notebook.
• dataset - folder with information provided to the students to carry out the experiments.
• report - evidence of written work.
• ComputerVisionGmmComplete2020FinSubmit.ipynb - the file containing partial code explained in the report mentioned.

To run the implementations located in src simply select the 'ComputerVisionGmmComplete2020FinSubmit.ipynb' file.

Note:

1. This jupyter notebook is considered part 1 of the project, containing code and explanation for the Gauissian Mixture modeling section.

2. After each numbered header for example 1.Dataset there will be a dedicated cell with code for the particular task .

3. This notebook takes aproximately 1-3 days to run on all the gmms necessary to infer since it works using the CPU.

4. Will include the code for the GMM section on Github some time later in the year 👍🏾

1. Computer Vision Lecturer @ Wits

2. Sbusiso Mgidi

3. Class of 2020 and 2021 @ Wits

4. X. Jin, P. Niu and L. Liu, ”A GMM-Based Segmentation Method for the Detection of Water Surface Floats,” in IEEE Access, vol. 7, pp. 119018-119025, 2019, doi: 10.1109/ACCESS.2019.2937129.

5. Farnoosh, R. and Zarpak, P.B., 2008. Image segmentation using Gaussian mixture model.

6. Github and google

• Repo owner Neil Fabião -> @neilfabiao ✌🏾