This paper presents a comprehensive survey of deep learning-based object pose estimation, covering instance-level, category-level, and unseen object pose estimation. It discusses various input data modalities, degrees of freedom (DoF) in output poses, object properties, and downstream tasks, providing a holistic understanding of the field. The survey also covers training paradigms, inference modes, application areas, evaluation metrics, and benchmark datasets, reporting the performance of state-of-the-art methods on these benchmarks to help readers select suitable methods for their applications. It highlights prevailing trends, their strengths and weaknesses, and identifies key challenges and promising research directions. The survey is organized into sections on datasets and metrics, instance-level methods, category-level methods, unseen object pose estimation methods, and applications. It discusses various datasets used for evaluating object pose estimation, including BOP Challenge datasets, rigid and articulated object datasets, and unseen methods datasets. The survey also covers evaluation metrics for 3DoF, 6DoF, 9DoF, and other metrics. It reviews instance-level methods, including correspondence-based, template-based, voting-based, and regression-based methods. It discusses category-level methods, including shape prior-based and shape prior-free methods, and unseen object pose estimation methods, including CAD model-based and manual reference view-based methods. The survey identifies key challenges in object pose estimation, such as reliance on labeled training data, difficulty in generalizing to novel objects, model compactness, and robustness in challenging scenarios. It also highlights trends in the evolution of object pose estimation techniques over the past decade and identifies promising research directions for future work. The survey provides a detailed overview of the field, including datasets, metrics, methods, and challenges, and aims to facilitate further research in this area.This paper presents a comprehensive survey of deep learning-based object pose estimation, covering instance-level, category-level, and unseen object pose estimation. It discusses various input data modalities, degrees of freedom (DoF) in output poses, object properties, and downstream tasks, providing a holistic understanding of the field. The survey also covers training paradigms, inference modes, application areas, evaluation metrics, and benchmark datasets, reporting the performance of state-of-the-art methods on these benchmarks to help readers select suitable methods for their applications. It highlights prevailing trends, their strengths and weaknesses, and identifies key challenges and promising research directions. The survey is organized into sections on datasets and metrics, instance-level methods, category-level methods, unseen object pose estimation methods, and applications. It discusses various datasets used for evaluating object pose estimation, including BOP Challenge datasets, rigid and articulated object datasets, and unseen methods datasets. The survey also covers evaluation metrics for 3DoF, 6DoF, 9DoF, and other metrics. It reviews instance-level methods, including correspondence-based, template-based, voting-based, and regression-based methods. It discusses category-level methods, including shape prior-based and shape prior-free methods, and unseen object pose estimation methods, including CAD model-based and manual reference view-based methods. The survey identifies key challenges in object pose estimation, such as reliance on labeled training data, difficulty in generalizing to novel objects, model compactness, and robustness in challenging scenarios. It also highlights trends in the evolution of object pose estimation techniques over the past decade and identifies promising research directions for future work. The survey provides a detailed overview of the field, including datasets, metrics, methods, and challenges, and aims to facilitate further research in this area.