/research/vision-for-manipulation/ Recent content in Vision for robotic manipulation on ViCoS Lab Hugo en-us /research/vision-for-manipulation/cedirnet-3dof/ Mon, 01 Jan 0001 00:00:00 +0000 /research/vision-for-manipulation/cedirnet-3dof/ <h1 id="cedirnet-3dof--cloth-grasping-in-3dof">CeDiRNet-3DoF — Cloth Grasping in 3DoF</h1> <p><strong>Winning method from the <a href="https://www.iri.upc.edu/groups/perception/ClothManipulationChallenge/">Cloth Manipulation and Perception Challenge at ICRA 2023</a></strong></p> <p>Object grasping is a core challenge in robotics and computer vision, and deformable objects like fabrics present additional complexity due to their non-rigid nature. We present <strong>CeDiRNet-3DoF</strong>, a deep learning model for grasp point detection that combines <strong>center direction regression</strong> with regression of the <strong>grasp pose angle</strong>. Our method achieved <strong>1st place in the perception task of the ICRA 2023 Cloth Manipulation Challenge</strong>, while outperforming state-of-the-art approaches in our benchmark, including transformer-based models.</p> /research/vision-for-manipulation/cedirnet-6dof/ Mon, 01 Jan 0001 00:00:00 +0000 /research/vision-for-manipulation/cedirnet-6dof/ <h2 id="vision-guided-robot-cloth-unfolding">Vision-guided robot cloth unfolding</h2> <p><strong>Second place at the <a href="https://airo.ugent.be/cloth_competition/">ICRA 2024 Cloth Competition</a>, outperforming nine of the ten other competing methods.</strong></p> <figure class="nested-video"> <video controls preload="metadata" playsinline poster="images/final-montage.jpg" aria-label="Examples of bimanual robot cloth unfolding using predicted 6-DoF grasp poses"> <source src="https://data.vicos.si/skokec/rtfm/CeDiRNet-6DoF/cedirnet_demo_h264_1080p_submission.mp4" type="video/mp4"> Your browser does not support embedded video. <a href="https://data.vicos.si/skokec/rtfm/CeDiRNet-6DoF/cedirnet_demo_h264_1080p_submission.mp4">Download the video</a>. </video> <figcaption>Examples of bimanual robot cloth unfolding using predicted 6-DoF grasp poses</figcaption> </figure> <h2 id="overview">Overview</h2> <p>Robotic cloth manipulation is difficult not only because fabric can assume an enormous range of configurations, but also because every grasp changes its geometry: folds move, layers separate, new self-occlusions appear, and visually similar grasp locations can produce very different outcomes. Unlike rigid objects, cloth has no stable shape from which a grasp can be planned directly. Successful unfolding therefore requires estimating both where to grasp and how to orient the gripper from a partial visual observation.</p>