SHIC: Shape-Image Correspondences with no Keypoint Supervision

TL;DR

SHIC introduces a novel, unsupervised approach to learning canonical surface mappings by leveraging foundation models and image generation, outperforming supervised methods across multiple categories.

Contribution

The paper presents SHIC, a method that learns shape-image correspondences without manual supervision by utilizing foundation models and image generation techniques.

Findings

Outperforms supervised methods on most categories

Leverages foundation models like DINO and Stable Diffusion

Uses image generation to enhance template realism

Abstract

Canonical surface mapping generalizes keypoint detection by assigning each pixel of an object to a corresponding point in a 3D template. Popularised by DensePose for the analysis of humans, authors have since attempted to apply the concept to more categories, but with limited success due to the high cost of manual supervision. In this work, we introduce SHIC, a method to learn canonical maps without manual supervision which achieves better results than supervised methods for most categories. Our idea is to leverage foundation computer vision models such as DINO and Stable Diffusion that are open-ended and thus possess excellent priors over natural categories. SHIC reduces the problem of estimating image-to-template correspondences to predicting image-to-image correspondences using features from the foundation models. The reduction works by matching images of the object to non-photorealistic renders of the template, which emulates the process of collecting manual annotations for this task. These correspondences are then used to supervise high-quality canonical maps for any object of interest. We also show that image generators can further improve the realism of the template views, which provide an additional source of supervision for the model.