TUD-L

Introduced 2018-08-24

The TUD-L (TUD Light) dataset is part of the Benchmark for 6D Object Pose Estimation (BOP). Let me provide you with some details about this dataset:

  1. Description:

    • The TUD-L dataset focuses on lighting conditions and includes three moving objects observed under eight different illumination scenarios.
    • It was specifically created as part of the BOP benchmark to evaluate 6D object pose estimation methods.

  2. Object Instances:

    • The dataset contains three objects:
      • Dragon
      • Frog
      • Watering pot

  3. Challenges:

    • TUD-L introduces challenging illumination variations to test the robustness of pose estimation algorithms.
    • Unlike other datasets, TUD-L has limited clutter and occlusion, making it suitable for evaluating lighting-specific performance.

  4. Ground-Truth Annotations:

    • The dataset provides ground-truth 6D object poses, 2D bounding boxes, and 2D binary masks for the training and test RGB-D images.
    • These annotations enable researchers to evaluate their algorithms accurately.

  5. Usage:

    • Researchers can use the TUD-L dataset to develop and evaluate 6D object pose estimation methods under varying lighting conditions.