Color segmentation of images using probabilistic methods

In this project I used Gaussian Mixture Models trained for the color red to segment color images. I used shape heuristics along with GMMs to detect red barrels in color images. Using ground truth on about 10 images, I also calibrated the size of the bounding box for distance of the barrel.

The final output is a bounding box around the barrel and an estimate of the distance of the barrel from the camera.

For more information have a look at the project report in the 'info' folder.