Contrast Based Saliency Filtering

1. Download dataset from here(109 MB)

2. Directory structure:

   1. src/main.py: Calculates the final saliency. Its major components are:

      1. apply_slic or apply_gmm
      2. apply_uniqueness
      3. apply_distribution
      4. apply_saliency

   2. src/slic_segmentation.py: Apply SLIC superpixels based on LAB color space and Euclidean distance using K-means clustering.

   3. src/gmm_segmentation.py: Fit GMM onto image pixels based on RGB, RGBXY and LABXY space. And divide pixels into n_components clusters to form Superpixels.

3. Usage:

   1. python src/main.py <path_to_image>
   2. python src/gmm_segmentation.py <path_to_image>
   3. Superpixels count can be varied by changing n_components in the code.

4. Resources:

   1. Saliency Filters(CVPR 2012)