CAR-API: Cityscapes Attributes Recognition API

This is the official api to download and fetch attributes annotations for Cityscapes Dataset.

Content

• Installation
• PyTorch Example
• Citation

Installation

You first need to download Cityscapes dataset. You can do so by checking this repo.

I'm showing here a simple working example to download the data but for further issues please refer to the source repo. Or download from the official website

1. Install Cityscapes scripts and other required packages.

$ pip install -r requirements.txt

2. Run the following script to download Cityscapes dataset. If you don't have an account, you will need to create an account.

$ csDownload -d [DESTINATION_PATH] PACKAGE_NAME

Note: you can also use -l option to list all possible packages to download. i.e.

$ csDownload -l

3. After downloading all required packages, set the environment variable CITYSCAPES_DATASET to the location of the dataset. For example, if the dataset is installed in the path /home/user/cityscapes/

$ export CITYSCAPES_DATASET="/home/user/cityscapes/"

Note: you can also export the previous command to your ~/.bashrc file for example.

$ echo 'export CITYSCAPES_DATASET="/home/user/cityscapes/"' > ~/.bashrc

Note2: we actually need the images only. We do not need the labels as it is stored with the attributes annotations as well.

4. Run the following to download the json files of CAR compressed as a single zip file extract it and then remove the zip file.

$ python download_CAR.py --url_path "https://DOWNLOAD_LINK_HERE"

To obtain the download link, please email me at kmetwaly511 [at] gmail [dot] com.

At this point, you have 4 json files; namely all.json, train.json, val.json and test.json

PyTorch Example

We provide a pytorch example to read the dataset and retrieve a sample of the dataset in pytorch_dataset_CAR.py. Please, refer to main.It contains a code that goes through the entire dataset.

An output sample of the dataset class is of custom type ModelInputItem. Please refer to the definiton of the class for more details about defined methods and variables.

Citation

If you are planning to use this code or the dataset, please cite the work appropriately as follows.

@misc{car_api,
  title = {{CAR}-{API}: an {API} for {CAR} Dataset},
  key = {{CAR}-{API}},
  howpublished = {\url{http://github.com/kareem-metwaly/car-api}},
  note = {Accessed: 2021-11-16}
}

@misc{metwaly2022car,
  title={{CAR} -- Cityscapes Attributes Recognition A Multi-category Attributes Dataset for Autonomous Vehicles}, 
  author={Kareem Metwaly and Aerin Kim and Elliot Branson and Vishal Monga},
  year={2021},
  eprint={2111.08243},
  archivePrefix={arXiv},
  primaryClass={cs.CV},
  howpublished = {\url{https://arxiv.org/abs/2111.08243}},
  urldate = {2021-11-17},
}