matteo-dunnhofer / fpv-tracking-baselines

Python implementation of the LTMU-H and TbyD-H trackers proposed in https://arxiv.org/abs/2209.13502

Implementation of FPV trackers

This repository contains the implementations of the tracking algorithms LTMU-H and TbyD-H presented in the paper "Visual Object Tracking in First Person Vision" appearing in the International Journal of Computer Vision (IJCV).

Refer to the respective README.md present in the LTMU-H and TbyD-H folders to know how to run each of the trackers.

Authors

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: [email protected]

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 = {2022}
}

@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}
}

License

All files in this dataset are copyright by us and published under the Creative Commons Attribution-NonCommercial 4.0 International License, found here. This means that you must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. You may not use the material for commercial purposes.

Copyright © Machine Learning and Perception Lab - University of Udine - 2021 - 2022

fpv-tracking-baselines's Issues

Can't run LTMU-H baseline evals for TREK-150: what's going on?

I'm testing out LTMU-H for TREK-150 and finding that even after setting up 100DoH and the LTMU setup + extracting and running TREK-150 (and verifying that the rest of the toolkit works), I'm unable to run evaluation for LTMU-H because the HIC detector python file doesn't seem to register the "model" package correctly. Basically the error is happening here:

import sys
sys.path.append('./hand_object_detector')
from model.utils.config import cfg, cfg_from_file, cfg_from_list, get_output_dir
from model.rpn.bbox_transform import clip_boxes
from model.roi_layers import nms
from model.rpn.bbox_transform import bbox_transform_inv
from model.utils.net_utils import filter_object, save_net, load_net, vis_detections, vis_detections_PIL, vis_detections_filtered_objects_PIL, vis_detections_filtered_objects # (1) here add a function to viz
from model.utils.blob import im_list_to_blob
from model.faster_rcnn.vgg16 import vgg16
from model.faster_rcnn.resnet import resnet

with the actual error being

Traceback (most recent call last):
  File "/home/vap43/fpv-tracking-baselines/LTMU-H/evaluate_trek150.py", line 4, in <module>
    from ltmuh import LTMUH
  File "/home/vap43/fpv-tracking-baselines/LTMU-H/ltmuh.py", line 30, in <module>
    from hic_detector import hic_config, hic_detect
  File "/home/vap43/fpv-tracking-baselines/LTMU-H/./hic_detector.py", line 13, in <module>
    from model.utils.config import cfg, cfg_from_file, cfg_from_list, get_output_dir
ModuleNotFoundError: No module named 'model.utils'; 'model' is not a package

Any idea what's going on? I'm looking to use this for a downstream task but am confused as to why the HIC is failing on this import: is there some setup issue?

OPE-D Evaluation Related Code?

Congratulations on the great work! I'm very interested in your work and evaluations. Would you be able to share the code described in the paper about OPE-D Evaluations? Thank you very much!

LTMU-H reproduction not performing as well as reported?

Hi there, thanks so much for the great work and toolkit for future benchmarks!

I'm running the LTMU-H baseline for TREK-150 under the OPE protocol to get some initial understanding of quantitative performance, and I'm finding that SS, NPS, and GSS are significantly lower than what's reported in the paper. I've posted my values below.

I followed the initial guidelines, so my initial thought is that there's something different between my setup and the setup used to run evaluation. Any idea as to what's going on?