Each image file in the dataset will have an associated XML file containing the ground-truth information i.e. the polygons and the labels of objects in the image file.
Input for participant’s methods
The dataset will be divided into training and test sets, to allow the participants to train and validate their methods.
For the training set the participants will be provided images (or links to floorplan images in case of images downloaded from Internet) and the corresponding XML files containing the ground-truth information i.e. polygons (or bounding boxes) and the labels of the objects in the corresponding image file.
For the test set the participants will be provided only the images (or the links to the floorplan images in case of images downloaded from Internet).
Output from participant’s methods
The participants will provide one XML file (containing label and positions of the detected objects) for each image in the test set.