Hand-2-Text

• 

  The letter Q in ASL (try it in our code if you want!) (it *will* fail)

• 

  The letter Y in ASL (try it in our code if you want!) (it ((((probably)))) won't fail).

Inspiration

I LOOOOOOVE NEURAL NETWORKS!!!! THEY'RE SO COOOOOOLLL!!!

What it does

not much

How we built it

We utilised TensorFlow and the Keras front end in order to develop our convolutional neural network. We utilised a Kaggle dataset containing over 87000 images of human hands signing the letters of the alphabet in order to train our CNN. For the interface, we decided to make a web app using the Django library for Python. This provided an easy way for us to quickly develop the simple website we required and integrate our CNN into the front end.

Challenges we ran into

Everything.

Accomplishments that we're proud of

Making the computer vision was really cool, but making it work reliably will be a task for next time!

What we learned

Our web-dev team learnt how to use Django and our data scientist team learnt how to use tensor flow as well as gaining an understanding of how CNNs function and why. We also all learnt that integrating multiple technologies is hard.

What's next for Hand-2-Text

Maybe getting it to actually work.

• Check out durhack.ipynb in our github repo to see the CNN actually working well! Before it broke somewhere in between putting it on the website...

Built With

• computer
• django
• github
• google-cloud
• keras
• programming
• python
• tensorflow