arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.01302cs.CVcs.AIcs.RO
收藏

面向抓取任务的少标注机器人实例分割

Robot Instance Segmentation with Few Annotations for Grasping

Technion – Israel Institute of Technology（以色列理工学院）；Bosch Center for AI（博世人工智能中心）；Ben-Gurion University of the Negev（内盖夫本-古里安大学）

Moshe Kimhi, David Vainshtein, Chaim Baskin, Dotan Di Castro

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对机器人操作中视觉感知依赖大量标注数据、泛化未知物体难的问题，提出结合半监督学习（SSL）与交互学习（LTI）的框架，通过观察场景变化和视觉一致性学习，无需交互序列数据，在两个基准上取得SOTA性能，ARMBench上AP₅₀达86.37且低标注场景表现优异。

摘要：

机器人的物体操作能力高度依赖其视觉感知能力。在杂乱场景和物体高变异性的领域中，大多数方法需要大量人工费力标注的数据集来训练高性能模型。部署后，模型泛化到未知物体的挑战意味着其必须随应用场景一同演进。为此，我们提出一种结合半监督学习（SSL）与交互学习（LTI）的新框架，使模型能通过观察场景变化学习，并利用视觉一致性（尽管存在时间间隔），无需交互序列的精心数据。因此，我们的方法通过自监督利用部分标注数据，并使用从无标注静态图像生成的伪序列整合时间上下文。我们在ARMBench mix-object-tote和OCID两个常见基准上验证了该方法，取得了最先进的性能。值得注意的是，在ARMBench上，我们获得了86.37的AP₅₀，比现有工作提高了近20%；在极低标注场景中，仅用1%的标注数据就获得了84.89的AP₅₀，而ARMBench全标注对应结果为72。

英文摘要：

The ability of robots to manipulate objects relies heavily on their aptitude for visual perception. In domains characterized by cluttered scenes and high object variability, most methods call for vast labeled datasets, laboriously hand-annotated, with the aim of training capable models. Once deployed, the challenge of generalizing to unfamiliar objects implies that the model must evolve alongside its domain. To address this, we propose a novel framework that combines Semi-Supervised Learning (SSL) with Learning Through Interaction (LTI), allowing a model to learn by observing scene alterations and leverage visual consistency despite temporal gaps without requiring curated data of interaction sequences. As a result, our approach exploits partially annotated data through self-supervision and incorporates temporal context using pseudo-sequences generated from unlabeled still images. We validate our method on two common benchmarks, ARMBench mix-object-tote and OCID, where it achieves state-of-the-art performance. Notably, on ARMBench, we attain an $\text{AP}_{50}$ of $86.37$, almost a $20\%$ improvement over existing work, and obtain remarkable results in scenarios with extremely low annotation, achieving an $\text{AP}_{50}$ score of $84.89$ with just $1 \%$ of annotated data compared to $72$ presented in ARMBench on the fully annotated counterpart.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑