Synthetic Data Generation and Vision-based Wrinkle and Keypoint Detection for Bimanual Cloth Manipulation
Researchers have developed a new method for robotic manipulation of textiles, addressing the challenge of visual perception in handling deformable fabrics. Their approach utilizes a synthetic data generation pipeline in Blender to create auto-annotated keypoints and combines this with real-world data to train a wrinkle detector. The system integrates a CNN for keypoint detection and a YOLOv8-OpenCV pipeline for identifying grasping points from wrinkles, enabling robots to stretch and iron garments effectively. AI
IMPACT Advances robotic perception for deformable objects, potentially improving automation in textile manufacturing and garment care.