This repository hosts an official toolkit for DTC, a high quality dataset for object reconstruction research.
The toolkit offers:
- A sequence download tool that will download user selected sequences and their corresponding 3D object models.
- Three example python codes illustrating how to load and use DTC dataset, including a Rerun visualizer for Aria sequence, object mask generators for Aria and DSLR sequences.
Resources:
The following are instructions to run the toolkit on DTC dataset.
The codebase is supported on:
- x64 Linux distributions of:
- Fedora 36, 37, 38
- Ubuntu jammy (22.04 LTS) and focal (20.04 LTS)
- Mac Intel or Mac ARM-based (M1) with MacOS 11 (Big Sur) or newer
Python 3.9+ (3.10+ if you are on Apple Silicon).
Install Python library dependencies in a virtual environment.
# 1. Create virtual environment
rm -rf $HOME/dtc_tools_python_env
python3 -m venv $HOME/dtc_tools_python_env
source $HOME/dtc_tools_python_env/bin/activate
# 2. Install Python packages
python3 -m pip install --upgrade pip
pip install -r ${PATH_TO_DTC_REPO}/requirements.txt
- Review the DTC license agreement.
- Examine the specific licenses applicable to the data types you wish to use, such as Sequence and 3D object models.
- Go to the DTC website and sign up.
- Scroll down to the bottom of the page.
- Enter your email and select Access the Datasets.
- The DTC page will be refreshed to contain instructions and download links
- The download view is ephemeral, keep the tab open to access instructions and links
- Download links that last for 14 days
- Enter your email again on the DTC main page to get fresh links
- Select the Download button for any of the data types:
- “Download the DTC Aria Dataset"
- "Download the DTC 3D Object Model Dataset"
- "Download the DTC DSLR Dataset"
- These will swiftly download JSON files with urls that the downloader will use
Use the Project Aria Tool downloader to download some, or all of the data.
# 1. Activate your environment
source $HOME/dtc_tools_python_env/bin/activate
# 2. Setup DTC Aria data folder
mkdir -p $HOME/Documents/dtc_aria
# 3. Run the dataset downloader
# Download all DTC Aria sequence data
aria_dataset_downloader -c ${PATH_TO_ARIA_CDN_FILE} -o $HOME/Documents/dtc_aria/ -d 0 1 2 3 6
# Type answer `y`
# Download one DTC Aria sequence data
aria_dataset_downloader -c ${PATH_TO_ARIA_CDN_FILE} -o $HOME/Documents/dtc_aria/ -l BirdHouseRedRoofYellowWindows_active -d 0 1 2 3 6
# Type answer `y`
# 4. Setup DTC DSLR data folder
mkdir -p $HOME/Documents/dtc_dslr
# 5. Run the dataset downloader
# Download all DTC DSLR sequence data
aria_dataset_downloader -c ${PATH_TO_DSLR_CDN_FILE} -o $HOME/Documents/dtc_dslr/ -d 0 1
# Type answer `y`
# Download one DTC DSLR sequence data
aria_dataset_downloader -c ${PATH_TO_DSLR_CDN_FILE} -o $HOME/Documents/dtc_dslr/ -l Airplane_B097C7SHJH_WhiteBlue_Lighting001 -d 0 1
# Type answer `y`
# 6. Setup DTC 3D object model data folder
mkdir -p $HOME/Documents/dtc_model
# 3. Run the dataset downloader
# Download all DTC 3D object model data
dtc_object_downloader -c {PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_model
# Type answer `y`
# Download one DTC 3D object model data
dtc_object_downloader -c {PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_model -l Airplane_B097C7SHJH_WhiteBlue
# Type answer `y`
# 1. Activate your environment
source $HOME/dtc_tools_python_env/bin/activate
# 2. Setup DTC Aria and 3D object model data folder
mkdir -p $HOME/Documents/dtc_aria_model
# 3. Run the dataset downloader
# Download all DTC Aria sequence with corresponding 3D object model data
python3 ${PATH_TO_DTC_REPO}/download_model_with_sequence.py -s ${PATH_TO_ARIA_CDN_FILE} -m ${PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_aria_model
# Type answer `y`
# Download one DTC Aria sequence with corresponding 3D object model data
python3 ${PATH_TO_DTC_REPO}/download_model_with_sequence.py -s ${PATH_TO_ARIA_CDN_FILE} -m ${PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_aria_model -l BirdHouseRedRoofYellowWindows_active
# Type answer `y`
# 4. Setup DTC DSLR and 3D object model data folder
mkdir -p $HOME/Documents/dtc_dslr_model
# 5. Run the dataset downloader
# Download all DTC DSLR sequence with corresponding 3D object model data
python3 ${PATH_TO_DTC_REPO}/download_model_with_sequence.py -s ${PATH_TO_DSLR_CDN_FILE} -m ${PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_dslr_model
# Type answer `y`
# Download one DTC DSLR sequence with corresponding 3D object model data
python3 ${PATH_TO_DTC_REPO}/download_model_with_sequence.py -s ${PATH_TO_DSLR_CDN_FILE} -m ${PATH_TO_OBJECT_CDN_FILE} -o $HOME/Documents/dtc_dslr_model -l BirdHouseRedRoofYellowWindows_active
# Type answer `y`
Viewing Aria rig trajectory, semi-dense point cloud and 3D object model
# 1. Activate your environment
source $HOME/dtc_tools_python_env/bin/activate
# 2. Run the dataset visualizer
python ${PATH_TO_DTC_REPO}/visualize_aria.py --sequence_folder $HOME/Documents/dtc_aria/{ARIA_SEQUENCE_NAME} --model_folder $HOME/Documents/dtc_model/{3D_MODEL_NAME}
# 1. Activate your environment
source $HOME/dtc_tools_python_env/bin/activate
# 2. Setup DTC Aria mask folder
mkdir -p $HOME/Documents/dtc_aria_mask
# 3. Run the mask generation
python ${PATH_TO_DTC_REPO}/generate_aria_mask.py --sequence_folder $HOME/Documents/dtc_aria/{ARIA_SEQUENCE_NAME} --model_folder $HOME/Documents/dtc_model/{3D_MODEL_NAME} --output_folder $HOME/Documents/dtc_aria_mask --width 800 --height 800 --focal 400
# 1. Activate your environment
source $HOME/dtc_tools_python_env/bin/activate
# 2. Setup DTC DSLR mask folder
mkdir -p $HOME/Documents/dtc_dslr_mask
# 3. Run the mask generation
python ${PATH_TO_DTC_REPO}/generate_dslr_mask.py --sequence_folder $HOME/Documents/dtc_aria/{ARIA_SEQUENCE_NAME} --model_folder $HOME/Documents/dtc_model/{3D_MODEL_NAME} --output_folder $HOME/Documents/dtc_dslr_mask
- DTC Toolkit (aka. this repository) is released by Meta under the Apache 2.0 license
- DTC dataset is released under the DTC license agreement
Go to Contributing and the Code of Conduct.