LM-O

LINEMOD-Occluded

The LM-O (Linemod-Occluded) dataset, introduced by Brachmann et al. in their work on 6D object pose estimation, provides additional ground-truth annotations for all modeled objects in one of the test sets from the Linemod (LM) dataset. This extension introduces challenging test cases with various levels of occlusion ¹²³.

Here are the key details about the LM-O dataset:

  • Objective: The primary goal of the LM-O dataset is to evaluate the performance of 6D object pose estimation methods under occlusion conditions.
  • Source: The dataset builds upon the Linemod dataset, which was originally proposed by Hinterstoisser et al. in their work on model-based training, detection, and pose estimation of texture-less 3D objects in heavily cluttered scenes ¹.
  • Annotations: LM-O provides additional ground-truth annotations for all other instances of the modeled objects in one of the Linemod test sets. These annotations include information related to occlusion, making it a valuable resource for assessing pose estimation algorithms in challenging scenarios ¹.
  • Test Cases: The LM-O dataset includes test images showing one annotated object instance with significant clutter but only mild occlusion. By incorporating occluded instances, it simulates real-world scenarios where objects may be partially hidden or obscured ¹.
  • License: The dataset is available under the CC BY-SA 4.0 license, allowing researchers to use and build upon it for their own investigations ¹.

Researchers and practitioners can leverage the LM-O dataset to develop and evaluate robust 6D object pose estimation methods that can handle occluded objects effectively. It serves as a valuable benchmark for advancing the field of computer vision and robotics.

(1) Datasets - BOP: Benchmark for 6D Object Pose Estimation. https://bop.felk.cvut.cz/datasets/. (2) BOP: Benchmark for 6D Object Pose Estimation - cvut.cz. https://cmp.felk.cvut.cz/~hodanto2/data/hodan2018bop_slides_eccv.pdf. (3) BOP: Benchmark for 6D Object Pose Estimation | SpringerLink. https://link.springer.com/chapter/10.1007/978-3-030-01249-6_2. (4) Recovering 6D Object Pose: A Review and Multi-modal Analysis. https://link.springer.com/chapter/10.1007/978-3-030-11024-6_2. (5) undefined. https://bop.felk.cvut.cz/media/data/bop_datasets.