DTTD-Mobile: Digital Twin Tracking Dataset Calibrated with a Optical MoCap System and a Mobile Device

🤔 Are current 3D object tracking methods truely robust enough for low-fidelity depth sensors like the iPhone LiDAR? We provide a new dataset on a mobile device: 18 objects observed in 100 videos with 47,668 sampled frames and 114,143 object annotations.

🌹 If our work is useful or relevant to your research, please kindly recognize our contributions by citing our papers:

@InProceedings{Huang_2025_CVPR,
    author    = {Huang, Zixun and Yao, Keling and Zhao, Zhihao and Pan, Chuanyu and Yang, Allen},
    title     = {Robust 6DoF Pose Estimation Against Depth Noise and a Comprehensive Evaluation on a Mobile Dataset},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops},
    month     = {June},
    year      = {2025},
    pages     = {1848-1857}
}

This repository contains an example dataset.py and simulated file structure within root. It is the dataset for the 2025 CVPRW paper Robust 6DoF Pose Estimation Against Depth Noise and a Comprehensive Evaluation on a Mobile Dataset.

🚩 Project page: https://openark-berkeley.github.io/DTTDNet/

🦾 Extension to Robotic Tasks: https://huggingface.co/datasets/ZixunH/DTTD3_Impedance

🏆 Leaderboard Embedded: https://paperswithcode.com/dataset/dttd2

🧸 You can also access our Transformer-based 6-DoF pose estimation network from: https://github.com/augcog/DTTD2

🍽️ And the checkpoints (trained on DTTD-Mobile): https://drive.google.com/drive/folders/1U7YJKSrlWOY5h2MJRc_cwJPkQ8600jbd?usp=share_link

📁 Folder Structure (including cad models, rgb images, semantic images, depth images, camera intrinsics, object poses, etc):

Robust-Digital-Twin-Tracking
├── checkpoints
│   ├── m8p4.pth
│   ├── m8p4_filter.pth
│   └── ...
|   ...
└── dataset
    └── dttd_iphone
        ├── dataset_config
        ├── dataset.py
        └── DTTD_IPhone_Dataset
            └── root
                ├── cameras
                │   ├── az_camera1 (if you want to train our algorithm with DTTD v1)
                │   ├── iphone14pro_camera1
                │   └── ZED2 (to be released...)
                ├── data
                │   ├── scene1
                │   │   └── data
                │   │   │   ├── 00001_color.jpg
                │   │   │   ├── 00001_depth.png
                │   │   │   └── ...
                |   │   └── scene_meta.yaml
                │   ├── scene2
                │   │   └── data
                |   │   └── scene_meta.yaml
                │   ...
                └── objects
                    ├── apple
                    │   ├── apple.mtl
                    │   ├── apple.obj
                    │   ├── front.xyz
                    │   ├── points.xyz
                    │   ├── ...
                    │   └── textured.obj.mtl
                    ├── black_expo_marker
                    └── ...

image image