Imitrob: Imitation Learning Dataset for Training and Evaluating 6D Object Pose Estimators

TL;DR

This paper presents Imitrob, a new dataset designed to improve 6D object pose estimation for hand-held tools in occluded, real-world task demonstrations, facilitating better training and evaluation of pose estimation methods.

Contribution

The paper introduces Imitrob, a comprehensive dataset with ground truth 6D poses for tools in manipulation tasks, addressing the lack of data for occluded object pose estimation in imitation learning.

Findings

Demonstrated use of the dataset by training and evaluating DOPE for 6D pose estimation.

Showed that the dataset enables robust training for occluded object scenarios.

Provided ground truth 6D poses for diverse manipulation tasks.

Abstract

This paper introduces a dataset for training and evaluating methods for 6D pose estimation of hand-held tools in task demonstrations captured by a standard RGB camera. Despite the significant progress of 6D pose estimation methods, their performance is usually limited for heavily occluded objects, which is a common case in imitation learning, where the object is typically partially occluded by the manipulating hand. Currently, there is a lack of datasets that would enable the development of robust 6D pose estimation methods for these conditions. To overcome this problem, we collect a new dataset (Imitrob) aimed at 6D pose estimation in imitation learning and other applications where a human holds a tool and performs a task. The dataset contains image sequences of nine different tools and twelve manipulation tasks with two camera viewpoints, four human subjects, and left/right hand. Each image is accompanied by an accurate ground truth measurement of the 6D object pose obtained by the HTC Vive motion tracking device. The use of the dataset is demonstrated by training and evaluating a recent 6D object pose estimation method (DOPE) in various setups.