Automated Accessibility Testing Tool

Crest quickly tests any HTML web page for a simple way to solve your accessibility testing problems. Provide a website URL to our callable API and expose potential WCAG violations.

Currently supports:

• Keyboard Focus Indicator
• Closed Captioning and Transcript
• Heading Analysis

Table of contents

• Getting Started
• Fine-tune Machine Learning Model
• Support
• Contributors
• Contributing
• License

Getting Started

To get you started quickly let’s dive into the necessary steps needed to set up the environment.

To use Crest as a Python Library

Using Pip

1. Install Crest using python package manager (Make sure pip is installed in your system).

pip install crest

To use Crest as a Service

Using Conda

1. Download and Install Miniconda from conda.io

2. For Windows, launch Miniconda prompt shell. For MacOS, launch the terminal.

3. Clone the Crest repository. More details can be found at docs.github.com

4. Go to "crest" folder in your terminal/prompt shell. Run command.

conda env create -f environment.yml
conda activate crest

5. Install chromedriver. The details can be found in chromium.org.

6. Download and Install google-chrome-stable from ubuntuupdates.org.

7. Once all the required libraries are installed, go to 'src/crest' folder and run the below command in your terminal/prompt shell.

FLASK_APP=server.py FLASK_ENV=development flask run --port 3000

Using Docker

1. Clone the repository. More details can be found at docs.github.com.

2. Download and Install docker from docker.com. Note: Make sure your docker server as well as crest image is running.

3. Go to "crest" folder in your terminal/prompt shell. Run command.

docker build -t crest .

4. Once the docker image is created, run it by executing the following command.

docker container run --name crest_container -p 3000:3000 crest 

Note: A machine learning model will be downloaded from huggingface.co when you use the Heading Analysis/ Crest Single API for the first time. It will be saved in your system's cache for future use and could take approximately 2 GB of your system's memory.

Fine-tune Machine Learning Model

Train the machine learning model to a specific domain:

1. Prepare a labelled dataset in the below format.

URL	Title	Description	Result
http://gov.uk	Tell us whether you accept cookies	We use cookies to collect information about how you use GOV.UK. We use this...	True

2. Change the dataset directory path i.e. ['model_params']['training_data_input_file_name'] in the config.py file.
3. Run a python script placed in the src/crest/utils/ directory by executing the following command:

    python train_model.py

Note: Install Apex if you are using fp16 training. Please follow the instructions in github.com/NVIDIA/apex. (Installing Apex from pip has caused issues for several people.)

Support

Crest is released as open source software and comes with no commercial support.
But since we want to ensure success and recognize that Crest consumers might fall into a range of roles - from developers that are steeped in the conventions of open-source to customers that are more accustomed to commercial offerings, we offer several methods of engaging with the Crest team and community.
For the Crest community, feel free to join our Slack channel

Contributors

Sheri Byrne Haber 💻

Amit Garg 💻

Joyce Oshita 💻

Contributing

The Crest project team welcomes contributions from the community. If you wish to contribute code, please look at our CONTRIBUTING.md file.

License

Crest is comprised of many open source software components, each of which has its own license that is located in the source code of the respective component as well as documented in the open source license file accompanying the Crest distribution.

---

The Crest team is inspired to make the web accessible to all. Join us to make it happen!