CNN_FACES

• 

  Model architecture shown in tensorboard

• 

  LinearSVC accuracy

Inspiration

Bio-metric authentication using emotions, this is two separate models used to determine the emotion of the person submitting a picture. Ideally it is combined with biometric authentication used to not only log in, but log in with a set expression to ensure another level of security.

What it does

This is a tensorflow convolutional net model to determine the mood of the user.

How I built it

I used a standard sized neural net due to the limitations of my hardware. But I have a VGG net built as well that could easily up the accuracy. Their is also a sense of ensemble methodology, I have also built a LinearSVC using Scikit-learn to make it's own prediction from here the two must agree to make a prediciton.

Challenges I ran into

Creating the front end for biometric login, I did this alone and didn't have time.

Accomplishments that I'm proud of

This is built entirely in tensorflow no higher level API.

What I learned

I learned more about multi-level classification there are 7 emotions to classify with this model

What's next for CNN_FACES

Implementing a user state can allow for more security authentication with facial biometric login.

Built With

• python