Starting Code Kit for the EPIC-KITCHENS TREK-150 Object Tracking Challenge

This repository shows how to run the LTMU-H tracker to obtain results that can be submitted to the Object Tracking CodaLab challenge.

The code presented here can be taken as baseline to implement and run your solution. The results to submit are obtained through the TREK-150-toolkit. To exploit the toolkit to obtain the results, all you need to do is to make your tracker inherit the class Tracker and override the init(self, image, box): and update(self, image): methods.

Challenge Organizers

Matteo Dunnhofer (1) Antonino Furnari (2) Giovanni Maria Farinella (2) Christian Micheloni (1)

(1) Machine Learning and Perception Lab, University of Udine, Italy
(2) Image Processing Laboratory, University of Catania, Italy

Contact: matteo.dunnhofer@uniud.it

Instructions to run the LTMU-H tracker

The following instructions demonstrate how to run the LTMU-H tracker defined in our IJCV paper.

Download and install LTMU-H

Download this repository

git clone https://github.com/matteo-dunnhofer/fpv-tracking-baselines
cd fpv-tracking-baselines/LTMU-H

Create the Conda environment and install the dependecies

conda env create -f environment.yml
pip install -f requirements.txt
conda activate ltmuh

Download the Hands-in-Contact repository
```
git clone https://github.com/ddshan/hand_object_detector.git
```
Download the pretrained model for egocentric data an put into a hand_object_detector/models/res101_handobj_100K/pascal_voc folder.
Download the LTMU repository
```
git clone https://github.com/Daikenan/LTMU.git
```
Set base_path = './LTMU' in the LTMU/DiMP_LTMU folder.

Download the STARK repository

git clone https://github.com/researchmm/Stark.git

Then run

python Stark/tracking/create_default_local_file.py --workspace_dir Stark/ --data_dir Stark/data --save_dir Stark/

Download the baseline pretrained model an put into a Stark/checkpoints/train/stark_st2 folder.

Download the TREK-150 dataset

Install the TREK-150 toolkit. Then, the full TREK-150 dataset can be built just by running

pip install got10k
git clone https://github.com/matteo-dunnhofer/TREK-150-toolkit
cd TREK-150-toolkit
python download.py

Run the tracker on the TREK-150 dataset

Yuo can run the LTMU-H tracker on TREK-150 by running the following command. The tracker will be run using the OPE, MSE, and HOI evaluation protocols.
```
cd ..
python evaluate_trek150_for_challenge.py
```

If you want to run your tracker, you have to make sure it inherits the class Tracker defined in Tracker and you have to override the init(self, image, box): and update(self, image): methods. The first is used to initialize the tracker in the first frame of the suquence, the second to get the tracker's predicted bounding-box in the other frames.

Prepare the results for the challenge

After having obained the results for each of the experimental protocols, you can create a zip file containing the valid JSON for the submission by running the following command

Run the evaluation on TREK-150 by running the following command.
```
python export_trek150_results_for_challenge.py
```

Implement and test your tracker

The exemplar scripts to implement your tracker according the toolkit and to run the experiments are given in the files

tracker.py
evaluate_trek150_for_challenge.py
export_trek150_results_for_challenge.py

of this repository.

Citing

When using the code, please reference:

@Article{TREK150ijcv,
author = {Dunnhofer, Matteo and Furnari, Antonino and Farinella, Giovanni Maria and Micheloni, Christian},
title = {Visual Object Tracking in First Person Vision},
journal = {International Journal of Computer Vision (IJCV)},
year = {2023}
}

@InProceedings{TREK150iccvw,
author = {Dunnhofer, Matteo and Furnari, Antonino and Farinella, Giovanni Maria and Micheloni, Christian},
title = {Is First Person Vision Challenging for Object Tracking?},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2021}
}

epic-kitchens/C8-Object-Tracking