Doduo: Learning Dense Visual Correspondence from Unsupervised Semantic-Aware Flow

TL;DR

Doduo is a self-supervised method that learns dense visual correspondence from in-the-wild images and videos, effectively handling dynamic scene changes for robotic perception tasks.

Contribution

Introduces Doduo, a novel self-supervised approach that incorporates semantic priors for robust dense correspondence learning without ground truth labels.

Findings

Outperforms existing self-supervised methods in point-level correspondence accuracy.

Effective in dynamic scenes with substantial transformations.

Demonstrates practical applications in robotics such as articulation estimation and zero-shot manipulation.

Abstract

Dense visual correspondence plays a vital role in robotic perception. This work focuses on establishing the dense correspondence between a pair of images that captures dynamic scenes undergoing substantial transformations. We introduce Doduo to learn general dense visual correspondence from in-the-wild images and videos without ground truth supervision. Given a pair of images, it estimates the dense flow field encoding the displacement of each pixel in one image to its corresponding pixel in the other image. Doduo uses flow-based warping to acquire supervisory signals for the training. Incorporating semantic priors with self-supervised flow training, Doduo produces accurate dense correspondence robust to the dynamic changes of the scenes. Trained on an in-the-wild video dataset, Doduo illustrates superior performance on point-level correspondence estimation over existing self-supervised correspondence learning baselines. We also apply Doduo to articulation estimation and zero-shot goal-conditioned manipulation, underlining its practical applications in robotics. Code and additional visualizations are available at https://ut-austin-rpl.github.io/Doduo