This repository serves as a comprehensive survey of deep learning-based object pose estimation, cataloging datasets, methods, and applications. It targets researchers and practitioners in computer vision and robotics seeking an overview of the field, providing a structured taxonomy and links to relevant papers and code.

How It Works

The survey categorizes object pose estimation into three main types: instance-level (specific objects), category-level (generalizing within a class), and unseen object pose estimation (generalizing to entirely new object categories). It further breaks down methods within these categories, such as correspondence-based, template-based, voting-based, and regression-based approaches, detailing their underlying principles and advancements.

Quick Start & Requirements

This repository is a survey and does not have a direct installation or execution command. It provides links to papers and code repositories for individual methods.

Highlighted Details

• Comprehensive categorization of object pose estimation methods.
• Extensive list of relevant datasets with chronological overview.
• Detailed breakdown of instance-level, category-level, and unseen object pose estimation techniques.
• Overview of applications across robotics, AR/VR, aerospace, and more.
• Links to papers and, where available, code for numerous research contributions.

Maintenance & Community

The repository is maintained by its authors, with an invitation for community contributions to add missing or recent papers.

Licensing & Compatibility

The repository itself is a collection of links and information; licensing depends on the individual linked projects.

Limitations & Caveats

As a survey, it reflects the state of research at the time of its last update and may not include the absolute latest advancements. The availability and quality of linked code repositories can vary.