Walking_Tours / README.md
shawshankvkt's picture
Update README.md
44bb99f verified
|
raw
history blame
2.51 kB
metadata
license: cc-by-4.0
task_categories:
  - image-classification
  - image-to-video
language:
  - en
tags:
  - self-supervised learning
  - representation learning
pretty_name: Walking_Tours
size_categories:
  - n<1K

Walking Tours Dataset

Overview

We introduce the Walking Tours dataset (WTours), a unique collection of long-range egocentric videos captured in an urban setting from various cities in Europe and Asia. It consists of 10 high-resolution videos, each showcasing a person walking through different urban environments, ranging from city centers to parks to residential areas under different lighting conditions. Additionally, a video from a Wildlife safari is included to diversify the dataset.

Cities Covered

The dataset encompasses walks through the following cities:

  • Amsterdam
  • Bangkok
  • Chiang Mai
  • Istanbul
  • Kuala Lumpur
  • Singapore
  • Stockholm
  • Venice
  • Zurich

Video Specifications

  • Resolution: 4K (3840 × 2160 pixels)
  • Frame Rate: 60 frames-per-second
  • License: Creative Commons License (CC-BY)

Duration

The videos vary in duration, offering a diverse range of content:

  • Minimum Duration: 59 minutes (Wildlife safari)
  • Maximum Duration: 2 hours 55 minutes (Bangkok)
  • Average Duration: 1 hour 38 minutes

Usage

The complete list of WTour videos are available in WTour.txt with YouTube link and the corresponding city.

To download the dataset, we first install pytube

pip install pytube

then, we run

python download_WTours.py --output_folder <path_to_folder> 

In order to comply with GDPR, we also try to blur out all faces and license plates appearing in the video using Deface

To do this for all videos in WTour dataset:

python3 -m pip install deface

Then run Deface on all videos using the bash script:

chmod a+x gdpr_blur_faces.sh  
./gdpr_blur_faces.sh

Citation

If you find this work useful and use it on your own research, please cite our paper:

@inproceedings{venkataramanan2023imagenet,  
  title={Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video},  
  author={Venkataramanan, Shashanka and Rizve, Mamshad Nayeem and Carreira, Jo{\~a}o and Asano, Yuki M and Avrithis, Yannis},  
  booktitle={International Conference on Learning Representations},  
  year={2024}  
}