CM_building_dataset_Timisoara

Date created: Febr-2020

The Timi\c{s}oara Building Dataset - TMBuD - is composed of 160 images with the resolution of 768x1024 pixels. Our motivation for this is the belief that this resolution is a good balance between the processing resources needed for manipulating the image and the actual resolution of pictures made with smart devices. Moreover, this is the actual video resolution for filming using a smartphone, the main sensor for building detection systems.

TMBuD is created from images of buildings in Timisoara. Each building ispresented from several perspectives, so this dataset can be used for evaluatinga building detection algorithm too. The dataset contains ground-truth imagesfor salient edges, for semantic segmentation and the GPS coordinates of thebuildings. The dataset contains 160 images grouped in the following sets: 100consist of the training dataset, 25 consist of the validation data and 35 consistof the test data.

This is the CM(Multimedia Center) building Image Dataset, which we created from images of buildings in Timisoara. The dataset contains groundthruth images for salient edges and semantic segmentation of building. Please check with the authors of the CM Building Dataset dataset, in case you are unsure about the respective copyrights and how they apply.

The dataset contains 160 images grouped in the following sets:

CLASSES :
BACKGROUND = 0 = ( 0, 0, 0)
BUILDING = 1 = ( 125, 125, 0)
DOOR = 2 = ( 0, 125, 125)
WINDOW = 3 = ( 0, 255, 255)
SKY = 4 = ( 255, 0, 0)
VEGETATION = 5 = ( 0, 255, 0)
GROUND = 6 = ( 125, 125, 125)
NOISE = 7 = ( 0, 0, 255)