Thanks to visit codestin.com
Credit goes to github.com

Skip to content
/ DVO Public
forked from princeton-vl/DPVO

DinoV3 Visual Odometry

License

HaktanM/DVO

 
 

Repository files navigation

Dino Visual Odometry

Setup and Installation

The code was tested on Ubuntu 24 and Cuda 12.

Clone the repo

git clone https://github.com/HaktanM/DVO.git
cd DVO

Create a virtual environment and activate it. Then

wget https://gitlab.com/libeigen/eigen/-/archive/3.4.0/eigen-3.4.0.zip
unzip eigen-3.4.0.zip -d thirdparty

# install
pip install torch --no-build-isolation
pip install -r requirements.txt --no-build-isolation
pip install . --no-build-isolation

# download models and data (~2GB)
./download_models_and_data.sh

Evaluation

EuRoC

Download all sequences from EuRoC (download ASL format)

python demo.py --eurocdir=<path to EuRoC> --calib=calib/euroc.txt

Training

Make sure you have run ./download_models_and_data.sh. Your directory structure should look as follows

├── datasets
    ├── TartanAir.pickle
    ├── TartanAir
        ├── abandonedfactory
        ├── abandonedfactory_night
        ├── ...
        ├── westerndesert
    ...

To train (log files will be written to runs/<your name>). Model will be run on the validation split every 10k iterations

python train.py --steps=240000 --lr=0.00008 --name=<your name>

Acknowledgements

  • The architecture is adopted from DPVO

About

DinoV3 Visual Odometry

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 73.5%
  • Python 15.8%
  • CMake 6.7%
  • Cuda 3.6%
  • Other 0.4%