SpecTrack: Learned Multi-Rotation Tracking via Speckle Imaging

Ziyang Chen, Mustafa Doğan, Josef Spjut and Kaan Akşit

[Website], [Manuscript]

Citation

@inproceedings{chen2024spectrack,
  author = {Ziyang Chen and Mustafa Dogan and Josef Spjut and Kaan Ak{\c{s}}it},
  title = {SpecTrack: Learned Multi-Rotation Tracking via Speckle Imaging},
  booktitle = {SIGGRAPH Asia 2024 Posters (SA Posters '24)},
  year = {2024},
  location = {Tokyo, Japan},
  publisher = {ACM},
  address = {New York, NY, USA},
  pages = {2},
  doi = {10.1145/3681756.3697875},
  url = {https://doi.org/10.1145/3681756.3697875},
  month = {December 03--06}
}

Getting started

(0) Requirments

Before using this code in this repository, please make sure to have the right dependencies installed. In order to install the main dependency used in this project, please make sure to use the below syntax in a Unix/Linux shell:

pip3 install -r requirements.txt

or

pip3 install git+https://github.com/kaanaksit/odak

or

git clone git@github.com:kaanaksit/odak.git
cd odak
pip install -e .

(1) Dataset

If training is needed, please download the training datasets from the URL and place the unzipped motion_dataset2_fft and static_dataset3+4_fftunder the dataset folder.

(2) Runtime

Once you have the main dependency installed, you can run the code base using the default settings by providing the below syntax:

git clone git@github.com:complight/SpecTrack.git
cd SpecTrack
python main.py --visual # with the speckle images showing in a window

If you want to train the model:

python train.py 

(3) Reconfiguring the code for your needs

Please consult the settings file found in settings/settings.txt, where you will find a list of self descriptive variables that you can modify according to your needs. This way, you can create a new settings file or modify the existing one. By typing,

python main.py --help

or

python train.py --help

You can reach to the information for training and estimating using this work.

If you are willing to use the code with another settings file, please use the following syntax:

python main.py --settings settings/sample.txt

or

python train.py --settings settings/sample.txt

Description

In this repository you will find the cadebase for the learn model discussed in our work. This work takes a captured laser speckle image from the lensless camera to estimate target surface rotatinal poses and depths. If you need support beyond these README.md files, please do not hesitate to reach us using issues section.

Support

For more support regarding the code base, please use the issues section of this repository to raise issues and questions or contact Ziyang.