Parsing World's Skylines using Shape Constrained MRFs

The above illustrates the input to the algorithm (left) and the parsing using rectangle MRF followed by refined MRF (right).

Bibtex

Below describes the code and data associated with the following paper. Consider citing this if you find it useful.

@inproceedings{tonge14CVPR,
    Author = {Rashmi Tonge, Subhransu Maji, and C.V. Jawahar},
    Title = {Parsing World's Skylines with Shape Constrained MRFs},
    Booktitle = {Computer Vision and Pattern Recognition},
    Year = {2014}
}

Link to the paper: skylines-cvpr14.pdf

Installation

Follow the README in the bitbucket repository and download the skyline 12 dataset from the link provided below.

skylineParsing github repository
skyline 12 dataset (~200 MB)

The code is released under the simplified BSD License (refer to the LICENSE file for details).

Demo

The code contains two demo files:

demoAnno.m: loads annotations and displays the ground truth parses for Chicago images. See comments for other examples.
demoParse.m: runs various methods and displays the outputs. Below is the output of the method (click on the image to zoom).