Model-free Approach to Garments Unfolding Based on Detection of Folded Layers

Jan Stria, Vladimir Petrik, Vaclav Hlavac

This page containts supplementary materials for our IROS 2017 paper.

Supplementary unfolding video

The supplementary video submitted with the paper. The first part of the vide shows our dual-arm robot unfolding a twice folded T-shirt. The second part shows unfolding of various garments.

Videos showing the experimental evaluation

The paper describes an experiment in which 4 different garments were unfolded in series of 5 uninterrupted trials. The following videos show uncut records of the experiment. The videos have been sped up 6 times. The T-shirt and towel were unfolded successfully in all 5 trials, the sweatshirt in 4 out of 5 trials and the shorts in 3 out of 5 trials.

Dataset of folded garments

The dataset used for evaluation can be downloaded here. It contains 13 different garments of 8 categories. Each garment was posed into 15 folded configurations, recorded and annotated manually. There are 6 files stored for each garment: RGB image, depth map, manually annotated folding axis, manually annotated polyline denoting the boundary of the top and bottom layer, found segmentation mask and labeling of pixels generated from the segmentation and the annotation.

Detailed experimental results

The detailed experimental results can be downloaded here. For each of 13 garments and its 15 folded configurations, we provide a single annotated image. It shows the contour of the garment found by the segmentation, the boundary of the detected top and bottom layer and the detected folding axis.

Presentation

You can download it here.