LabelMeFacade

Data created September/3/2010

Authors: Bjoern Froehlich and Erik Rodner

What is this?

A GUI program to help users quickly label unconstrained photographs of city scenes, like this:

Using the mouse to quickly label the image using the LabelMe tool, Result this this image plus target data row serializing this to target supervisory signal:

You could use this labelled image data to train a machine learning algorithm to find windows, buildings, roads, cars, shrubberies, etc.

Label classification color legend:

The dataset under images/ contains 100 images for training and 845 images for testing (see train.txt and test.txt for details). Color codes for labels are (in R:G:B):

various = 0:0:0
building = 128:0:0
car = 128:0:128
door = 128:128:0
pavement = 128:128:128
road = 128:64:0
sky = 0:128:128
vegetation = 0:128:0
window = 0:0:128

Explanation

This is the LabelMeFacade Image Dataset, which we created from LabelMe images for semantic segmentation research. Since this is a subset of LabelMe images, the images were originally collected by the authors of

LabelMe: A Database and Web-based Tool for Image Annotation Bryan C. Russell, Antonio Torralba, Kevin P. Murphy,William T. Freeman: http://publications.csail.mit.edu/tmp/MIT-CSAIL-TR-2005-056.pdf .

All images should only be used for non-commercial and research experiments. Please check with the authors of the LabelMe dataset, in case you are unsure about the respective copyrights and how they apply.

Citation

If you use this database please cite one of the following papers:

@INPROCEEDINGS{Froehlich-Rodner-Denzler-ICPR2010,
    author = {Bj{\"o}rn Fr{\"o}hlich and Erik Rodner and Joachim Denzler},
	title = {A Fast Approach for Pixelwise Labeling of Facade Images},
    booktitle = {Proceedings of the International Conference on Pattern Recognition
	(ICPR 2010)},
	year = {2010},
}

@inproceedings{Brust15:ECP,
    author = {Clemens-Alexander Brust and Sven Sickert and Marcel Simon and Erik Rodner and Joachim Denzler},
    booktitle = {CVPR Workshop on Scene Understanding (CVPR-WS)},
    title = {Efficient Convolutional Patch Networks for Scene Understanding},
    year = {2015},
}