CTVG

These grasps represent good points to manipulating the fabric without it losing its form, crucial to taking the parts to a sewing machine, for example. Considering a task where a robot simply has to pick up a piece of fabric and move it to somewhere else, the robot has to place its grippers on key areas of the object (grasping points) in such a way that when it goes to release the textile down on a surface, it is able to properly maintain the same 2D form as before picking it up.

For that, this dataset was created, containing information about where the grasping points should be for many different types of cut pieces. The dataset images have a resolution of 448x448 and are in .jpeg format. There are 4 types of images on this dataset: - full: synthetic images with all the cut pieces of each garment per image and for each of the 6 distinct textile patterns; - single: synthetic images with a single cut piece per image, for all cut pieces of each garment and for each of the 6 distinct textile patterns; - binary: synthetic images with a single cut piece per image, for all cut pieces of each garment, without any texture (object is white and background is black); - real: real images for all garments with only 3 distinct textile patterns, where the photos were taken with a camera Intel Realsense 435.