📱 TMO Dataset

Welcome to our dataset! This collection was created using our custom smartphone application running on an iPhone 13 Pro with ARKit 6. The app captures a variety of rich sensor outputs:

• 🖼️ Synchronized RGB images
• 🧠 Confidence maps
• 🌊 Dense depth maps from the LiDAR scanner
• 📈 IMU sensor data
• 📸 SfM results from RGBD-SfM
• 📍 Camera pose trajectory (ARKit)

We provide all raw data from ARKit.

You can download the dataset from our Google Drive.

---

🗂️ Dataset Types

We offer two types of datasets, tailored for different use cases:

1. ARKit-Video 🎬

• Captures long video sequences similar to RGBD-SLAM datasets
• 📸 Thousands of frames per sequence
• 🖥️ Resolution: 1280x720
• 📁 Each scene contains the following folders:
  • cameraParam/: Camera parameters and calibration data
  • confidence/: Confidence maps
  • depth/: Raw depth maps
  • gravity/: Gravity
  • rgb/: Original RGB images
  • rgb_keyframe/: Selected keyframe RGB images
  • rgbdsfm/sfm_final: Structure from Motion data
  • trajectory/: Camera pose trajectory
  • filtered_depth/: Processed depth maps
  • filtered_keyframe_depth/: Filtered depth maps for keyframes
  • preprocessed/: Preprocessing data for training neural surface reconstruction methods

2. AR-Capture 🎞️

• A multi-view object dataset created using Apple's Object Capture API
• 📷 High-resolution 4K images with depth and gravity data
• 🧮 Usually fewer than 50 images per object
• 🖼️ Image resolution: 4032x3096
• 📁 Each scene contains the following folders:
  • cameraParam/: Camera parameters and calibration data
  • confidence/: Confidence maps
  • depth/: Raw depth maps
  • gravity/: Gravity
  • rgb/: Original RGB images
  • rgb_keyframe/: Selected keyframe RGB images
  • rgbdsfm/sfm_final: Structure from Motion data
  • trajectory/: Camera pose trajectory
  • filtered_depth/: Processed depth maps
  • filtered_keyframe_depth/: Filtered depth maps for keyframes
  • preprocessed/: Preprocessing data for training neural surface reconstruction methods

🧸 Object Categories

We currently provide data for the following 11 objects:

• 🤖 Robot Arm 2
• 🌿 Plant
• 🌳 Tree
• 🛋️ Sofa
• 🚲 Bike
• 🪑 Recliner Chair
• ☕ Cafe Stand
• 🪑 Office Chair
• 🎥 Camera Stand
• 👟 Shoe
• 🤖 Delivery Robot

---

We hope this dataset helps you build amazing AR/AI applications! ✨
Feel free to reach out or contribute if you're using it 💡💬

📚 Citation

If you use this dataset in your research, please cite our paper:

@inproceedings{choi2023tmo,
    title = {TMO: Textured Mesh Acquisition of Objects with a Mobile Device by using Differentiable Rendering},
    author = {Jaehoon Choi and Dongki Jung and Taejae Lee and Sangwook Kim and Youngdon Jung and Dinesh Manocha and Donghwan Lee},
    booktitle = {CVPR},
    year = {2023}
}

---