Object Pose Estimation Using Implicit Representation for Transparent Objects

This dataset aggregates high quality 3D mesh assets and rendered data for training and fine-tuning pose estimation models. It unifies four significant datasets: ClearPose, DIMO, HouseCat6D, and TRansPose, all formatted for BOP (Benchmark for 6D Object Pose Estimation) evaluation.

Visualization

Below is a collage showing sample RGB inputs from the constituent datasets:

Component Datasets

This collection aggregates four datasets converted to the standard BOP format to facilitate comparative evaluation.

1. ClearPose

Source: ClearPose Project ClearPose consists of 63 transparent or opaque household objects captured in various lighting conditions and occlusions (e.g., glass utensils on a tabletop).

• Note: This dataset includes the downsampled version (every 100th image from each scene) to account for the small pose differences between consecutive frames in continuous motion and to decrease inference time.

2. DIMO (Dataset of Industrial Metal Objects)

Source: DIMO Project DIMO consists of six reflective metallic parts (colored, shiny, and matte finish) on a metallic surface. These objects exhibit a shiny appearance, designed to challenge rendering and pose estimation methods that must model view dependent effects.

3. HouseCat6D

Source: HouseCat6D Project A large scale multi modal dataset consisting of textured, shiny, metallic, and matte objects of different categories in realistic scenarios. It serves as a comprehensive benchmark for diverse household objects.

4. TRansPose

Source: TRansPose Project TRansPose consists of 99 transparent objects (glassy and plastic objects with different optical properties) cluttered on a tabletop.

• Note: Similar to ClearPose, this dataset is downsampled (every 10th image) from the original continuous motion capture. While the original setup included RGB, RGBD, and TIR images, this distribution relies on RGB images and bounding boxes, as the method requires only RGB and 2D detection.

Visualization

Randomly selected images from these datasets can be seen below:

• (a) ClearPose: Glass utensils on a tabletop.
• (b) DIMO: Shiny metallic parts.
• (c) HouseCat6D: Diverse textured/matte objects.
• (d) TRansPose: Cluttered glassy/plastic objects.

Dataset Structure

The dataset is organized into zipped archives for access.

Archives

• ClearPose.zip
• DIMO.zip
• HouseCat6D.zip
• TRansPose.zip

Internal File Structure (after unzipping)

Each sub-dataset follows the BOP directory structure:

dataset/
├── ClearPose/
│   ├── models/                 # Object models (.ply, .obj, models_info.json)
│   ├── nerfs/                  # NeRF data (Instant NGP format)
│   ├── test/                   # Test scenes
│   │   ├── 001001/             # Scene ID
│   │   │   ├── scene_camera.json # Camera parameters
│   │   │   ├── scene_gt.json     # Ground truth 6D poses
│   │   │   ├── rgb/            # RGB Images
│   │   │   ├── depth/          # Depth Maps
│   │   │   └── mask_visib/     # Visibility masks
│   │   └── ...
│   └── test_targets_bop19.json # BOP-style test targets
├── DIMO/
│   └── ...
├── HouseCat6D/
│   └── ...
└── TRansPose/
    └── ...

This work is associated with the research published in Object Pose Estimation Using Implicit Representation for Transparent Objects, available at SpringerView.

In creating this dataset, we utilized the BOP Toolkit for standardized formatting. This work was supported by the European Union under the project Robotics and advanced industrial production (reg. no. CZ.02.01.01/00/22_008/0004590).

The dataset is distributed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Users are free to share and adapt the material provided they give appropriate credit to the authors and the associated publication.

If you find this dataset useful, please cite it using:

@InProceedings{10.1007/978-3-031-91569-7_15,
author="Burde, Varun and Moroz, Artem and Zeman, V{\'i}t and Burget, Pavel",
editor="Del Bue, Alessio and Canton, Cristian and Pont-Tuset, Jordi and Tommasi, Tatiana",
title="Object Pose Estimation Using Implicit Representation for Transparent Objects",
booktitle="Computer Vision -- ECCV 2024 Workshops",
year="2025",
publisher="Springer Nature Switzerland",
address="Cham",
pages="226--247",
isbn="978-3-031-91569-7"
}