Nowadays, security is only as good as you make it. The PDFs we have today can be signed by just about anyone without any real sort of evidence that they were in fact the person who signed it. How can we really know if documents were signed by Joseph Song or Cassidy Williams? With Sincerely, we're able to train signature models to be able to accurately and verifiably predict whether it was the person who signed it or not.

What it does

Sincerely uses Wacom's Will SDK to take a signature from a user, saves it as an image on imgur, and uses Clarifai's machine learning technology to train and store the model. Once this information is stored, anyone can then use Sincerely to securely sign a document.

How I built it

We thought that adding machine learning to signature verification would be a good idea so we thought of how we could demo the product. This lead to creating a couple of html templates and some javascript functions.

Challenges I ran into

Initially, the first hurdle we had to overcome was Wacom's SDK and trying to run the samples. But once that was done, we found the Wacom, although could still use a few improvements, had much to offer.

+ 22 more
Share this project:


posted an update

If you'd like to run our application, head over to our GitHub repo here:

And then follow these steps to run the server:

cd sincerely
python -m SimpleHTTPServer

Open up localhost and check it out! Add your name (or whoever's name) to the box and click "Submit Name". When you do this, that creates a concept of your name. Then, draw your signature in the canvas below, and click "Train Signature". You can right click to clear the canvas, and continue doing this to your heart's content.

Now, if you want to test your signature, have a friend (or use your other hand) to try writing your signature again, and hit "Test Signature". The system will then tell you if your signature is verified by our machine learning system, powered by Clarifai and WILL. Enjoy!

Log in or sign up for Devpost to join the conversation.