LabelMe语义分割数据集

数据集包括用于训练的100张图像和用于测试的845张图像（详细信息请参阅 train.txt 和 test.txt ）。 标签的颜色代码为（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

数据引用：

@INPROCEEDINGS{Froehlich-Rodner-Denzler-ICPR2010,    
	author = {Bj"orn Fr"ohlich 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}, }