Multi-Aspect Reconstruction and Multi-Object Tracking

This repository contains all the steps necessary to tackle the multi-view multi-people tracking problem. The pipeline consists of 5 steps: calibration, annotation, training, inference and tracking.

System setup

Pulling the Repository

Due to the included submodule please use: git clone --recurse-submodules https://github.com/cvlab-epfl/MARMOT.git

Alternatively, clone and: git submodule update --init --recursive

More information about system setup can be found in the setup Readme.

Running the pipeline

Each step of the pipeline is describe in detail in documentation.

• 0-calibration
• 1-annotation
• 2-training
• 3-inference
• 4-tracking

Reference

If you found this code useful, please cite us:

@misc{MARMOT2023,
author        = {Engilberge, Martin and Grosche, Wilke and Fua, Pascal},
year          = {2023},
title         = {Multi-Aspect Reconstruction and Multi-Object Tracking},
howpublished = {\url{https://github.com/wgrosche/MARMOT/}}
}

@inproceedings{engilber2023multi,
  title={Multi-view Tracking Using Weakly Supervised Human Motion Prediction},
  author={Engilberge, Martin and Liu, Weizhe and Fua, Pascal},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  year={2023}
}

@inproceedings{engilber2023two,
    title={Two-level Data Augmentation for Calibrated Multi-view Detection},
    author={Engilberge, Martin and Shi, Haixin and Wang, Zhiye and Fua, Pascal},
    booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
    year={2023}
}

The annotation tool is based on the following work:

• Multicam-Gt

License

By downloading this program, you commit to comply with the license as stated in the LICENSE file.