Box Pose and Shape Estimation and Domain Adaptation for Large-Scale Warehouse Automation

2025-07-01

AI Summary

Addressing the scarcity of real-world annotated data for box pose and shape estimation in warehouse automation, this paper proposes a “correct-and-verify” self-supervised domain adaptation framework enabling zero-shot simulation-to-reality transfer. Without human annotations, the method jointly optimizes domain alignment and 3D bounding box regression via self-supervised learning. Starting from a simulation-pretrained model, it performs self-supervised fine-tuning on 50,000 unlabeled real images. Its core innovation is a learnable geometric consistency verification mechanism that dynamically refines pseudo-labels and mitigates domain shift. Evaluated on a large-scale industrial real-world dataset, the approach significantly outperforms both pure simulation-trained and zero-shot baseline methods, achieving a 12.6% improvement in 3D detection accuracy (AP₅₀). This work establishes a novel paradigm for robust pose estimation in annotation-free scenarios.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Sampling/Simulation-based Search
• Intelligent Robots: State Estimation

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Modern warehouse automation systems rely on fleets of intelligent robots that generate vast amounts of data -- most of which remains unannotated. This paper develops a self-supervised domain adaptation pipeline that leverages real-world, unlabeled data to improve perception models without requiring manual annotations. Our work focuses specifically on estimating the pose and shape of boxes and presents a correct-and-certify pipeline for self-supervised box pose and shape estimation. We extensively evaluate our approach across a range of simulated and real industrial settings, including adaptation to a large-scale real-world dataset of 50,000 images. The self-supervised model significantly outperforms models trained solely in simulation and shows substantial improvements over a zero-shot 3D bounding box estimation baseline.

Problem

• Self-supervised domain adaptation for warehouse robots
• Estimating box pose and shape without manual annotations
• Improving perception models using unlabeled real-world data

Innovation

• Self-supervised domain adaptation pipeline
• Correct-and-certify pose estimation
• Leverages unlabeled real-world data