Most websites fail to consider the necessary accommodations for disabilities in their web design, making the internet unfairly difficult to navigate for disabled people. Static font sizes on web pages make it difficult to read the contents. Images without alt-text make the web a harder place to navigate with a screen reader. What are ways the web is made inaccessible to people with disabilities who need screen readers or have visual impairments? TiA aims to solve this issue.

What it does

TiA is a chrome extension that attempts to make the internet easier and safer to navigate for people with visual impairments.

  • Change the Font Size for texts on a webpage.
  • Check if images on a website have alt-text.
  • Incorporate AI that can analyze the contents of the image and auto-generate alt-text.
  • Place alt-text in the empty spaces for the images on the webpage.

How we built it

We used Javascript, JSON, and HTML/CSS to create the chrome extension, which we tested on various browser websites. We used web API to capture content from the active tab and send the data to the extension popup. The popup then sends the image to be processed as a message to the server host, from which the python program accesses the URL. A deep learning model is incorporated into the extension as it inputs an image and returns an automatic caption. To accomplish this, we figured that we would need to use a Flickr Image Dataset with 13k photos to train a model. We cut down the dataset so that we could run the program locally and used different libraries like PyTorch to fit and build our model. We chose to employ the CNN and RNN neural networks for encoding and decoding and trained the model in 6 epochs.

Challenges we ran into

Since we had little previous experience working with Javascript and JSON, one challenge was learning the languages and APIs as we create the product. Additionally, none of us have had any experience working with Deep Learning whatsoever. Not only was it hard to run the program locally, as it took up too much memory and space on our computers, but it was also difficult to figure out. We ran into a lot of errors that we had to fix line by line.

Accomplishments that we're proud of

We're proud of creating a product that may help make the internet more accessible, especially since it incorporates aspects that we've all never worked with. We learned how to work with chrome extensions and deep learning and it was a very exciting process.

What we learned

We learned a lot of new computer science concepts, along with the fact that we can work with extremely hard libraries and languages.

What's next for TiA: Text & Image Accessibility

We hope to be able to extend TiA to make websites even more accessible, like changing website text/background colors to make high contrast visibility. We also would like to have a greater connection between the different aspects of the project, like the deep learning model.

Built With

Share this project: