Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Adrian, David B., Kupcsik, Andras Gabor, Spies, Markus, Neumann, Heiko
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2406.12441
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913396550533120
author Adrian, David B.
Kupcsik, Andras Gabor
Spies, Markus
Neumann, Heiko
author_facet Adrian, David B.
Kupcsik, Andras Gabor
Spies, Markus
Neumann, Heiko
contents Robot manipulation relying on learned object-centric descriptors became popular in recent years. Visual descriptors can easily describe manipulation task objectives, they can be learned efficiently using self-supervision, and they can encode actuated and even non-rigid objects. However, learning robust, view-invariant keypoints in a self-supervised approach requires a meticulous data collection approach involving precise calibration and expert supervision. In this paper we introduce Cycle-Correspondence Loss (CCL) for view-invariant dense descriptor learning, which adopts the concept of cycle-consistency, enabling a simple data collection pipeline and training on unpaired RGB camera views. The key idea is to autonomously detect valid pixel correspondences by attempting to use a prediction over a new image to predict the original pixel in the original image, while scaling error terms based on the estimated confidence. Our evaluation shows that we outperform other self-supervised RGB-only methods, and approach performance of supervised methods, both with respect to keypoint tracking as well as for a robot grasping downstream task.
format Preprint
id arxiv_https___arxiv_org_abs_2406_12441
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Cycle-Correspondence Loss: Learning Dense View-Invariant Visual Features from Unlabeled and Unordered RGB Images
Adrian, David B.
Kupcsik, Andras Gabor
Spies, Markus
Neumann, Heiko
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Robot manipulation relying on learned object-centric descriptors became popular in recent years. Visual descriptors can easily describe manipulation task objectives, they can be learned efficiently using self-supervision, and they can encode actuated and even non-rigid objects. However, learning robust, view-invariant keypoints in a self-supervised approach requires a meticulous data collection approach involving precise calibration and expert supervision. In this paper we introduce Cycle-Correspondence Loss (CCL) for view-invariant dense descriptor learning, which adopts the concept of cycle-consistency, enabling a simple data collection pipeline and training on unpaired RGB camera views. The key idea is to autonomously detect valid pixel correspondences by attempting to use a prediction over a new image to predict the original pixel in the original image, while scaling error terms based on the estimated confidence. Our evaluation shows that we outperform other self-supervised RGB-only methods, and approach performance of supervised methods, both with respect to keypoint tracking as well as for a robot grasping downstream task.
title Cycle-Correspondence Loss: Learning Dense View-Invariant Visual Features from Unlabeled and Unordered RGB Images
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2406.12441