KeyPointDiffuser: Unsupervised 3D Keypoint Learning via Latent Diffusion Models
This addresses the problem of representing 3D object structure for computer vision and graphics, bridging unsupervised keypoint learning with generative models.
The paper tackles unsupervised learning of 3D keypoints from point cloud data, achieving a 6 percentage-point improvement in keypoint consistency over prior methods.
Understanding and representing the structure of 3D objects in an unsupervised manner remains a core challenge in computer vision and graphics. Most existing unsupervised keypoint methods are not designed for unconditional generative settings, restricting their use in modern 3D generative pipelines; our formulation explicitly bridges this gap. We present an unsupervised framework for learning spatially structured 3D keypoints from point cloud data. These keypoints serve as a compact and interpretable representation that conditions an Elucidated Diffusion Model (EDM) to reconstruct the full shape. The learned keypoints exhibit repeatable spatial structure across object instances and support smooth interpolation in keypoint space, indicating that they capture geometric variation. Our method achieves strong performance across diverse object categories, yielding a 6 percentage-point improvement in keypoint consistency compared to prior approaches.