Category Level 6D Object Pose

Category-level 6D object pose estimation aims to determine an object's 3D position and orientation from a single image, even for unseen instances within a known category, overcoming limitations of instance-level methods. Current research focuses on improving generalization across diverse object shapes and viewpoints, employing techniques like keypoint learning, novel view synthesis using diffusion models, and two-stage deformation-and-registration pipelines, often leveraging shape priors or self-supervised learning to reduce reliance on large annotated datasets. This field is crucial for robotics and augmented reality applications requiring robust object manipulation and interaction in unstructured environments, driving advancements in both model architectures and benchmark datasets.