Skip to content

grasp-lyrl/dpvo_features

 
 

Repository files navigation

Deep Patch Visual Odometry

This repository contains the source code for our paper:

Deep Patch Visual Odometry
Zachary Teed, Lahav Lipson, Jia Deng

@article{teed2023deep,
  title={Deep Patch Visual Odometry},
  author={Teed, Zachary and Lipson, Lahav and Deng, Jia},
  journal={Advances in Neural Information Processing Systems},
  year={2023}
}

Setup and Installation

The code was tested on Ubuntu 20 and Cuda 11.
Update 9/12: We have an official Docker

Clone the repo

git clone https://github.com/princeton-vl/DPVO.git --recursive
cd DPVO

Create and activate the dpvo anaconda environment

conda env create -f environment.yml
conda activate dpvo

Next install the DPVO package

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

# install DPVO
pip install .

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

Recommended - Install the Pangolin Viewer

Note: You will need to have CUDA 11 and CuDNN installed on your system.

  1. Step 1: Install Pangolin (need the custom version included with the repo)
./Pangolin/scripts/install_prerequisites.sh recommended
mkdir Pangolin/build && cd Pangolin/build
cmake ..
make -j8
sudo make install
cd ../..
  1. Step 2: Install the viewer
pip install ./DPViewer

For installation issues, our Docker Image supports the visualizer.

Demos

DPVO can be run on any video or image directory with a single command. Note you will need to have installed DPViewer to visualize the reconstructions. The pretrained models can be downloaded from google drive models.zip if you have not already run the download script.

python demo.py \
    --imagedir=<path to image directory or video> \
    --calib=<path to calibration file> \
    --viz # enable visualization
    --plot # save trajectory plot
    --save_reconstruction # save point cloud as a .ply file
    --save_trajectory # save the predicted trajectory as .txt in TUM format

iPhone

python demo.py --imagedir=movies/IMG_0492.MOV --calib=calib/iphone.txt --stride=5 --plot --viz

TartanAir

Download a sequence from TartanAir (several samples are availabe from download directly from the webpage)

python demo.py --imagedir=<path to image_left> --calib=calib/tartan.txt --stride=1 --plot --viz

EuRoC

Download a sequence from EuRoC (download ASL format)

python demo.py --imagedir=<path to mav0/cam0/data/> --calib=calib/euroc.txt --stride=2 --plot --viz

Evaluation

We provide evaluation scripts for TartanAir, EuRoC, TUM-RGBD and ICL-NUIM. Up to date result logs on these datasets can be found in the logs directory.

TartanAir:

Results on the validation split and test set can be obtained with the command:

python evaluate_tartan.py --trials=5 --split=validation --plot --save_trajectory

EuRoC:

python evaluate_euroc.py --trials=5 --plot --save_trajectory

TUM-RGBD:

python evaluate_tum.py --trials=5 --plot --save_trajectory

ICL-NUIM:

python evaluate_icl_nuim.py --trials=5 --plot --save_trajectory

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>

Change Log

  • Aug 08, 2022: Initial release
  • Sep 12, 2022: Add link to docker
  • Mar 04, 2023: Google Colab, TUM + ICL-NUIM evaluation code, flags for saving output

Acknowledgements

  • Our Viewer is adapted from DSO.

About

Deep Patch Visual Odometry

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 67.3%
  • Python 21.4%
  • CMake 5.8%
  • Cuda 5.1%
  • Other 0.4%