MedAI

The future of mental health diagnoses. MedAI uses a machine learning model to predict mood of consumers.

Comment

Medai works by using neural networks to learn how to recognize human emotions. Every emotion is a biochemical response, which in turn causes an external, physical change in one's facial details. These details are captured in a picture.

The picture becomes the input for our neural network, trained on a diverse dataset of thousands of images of people expressing emotion, both explicitly and subtly. The network classifies the input image and returns a probability distribution of the subject's current emotional state.

Inspiration

The importance of empathy, the awareness of depression, as well as what we can do to improve the state of both human conditions. With the growing importance of mental health, we found it extremely important to develop an application that makes it easier for physicians to diagnose patients who may have mental health disorders.

What it does

MedAI uses machine learning to accurately predict mental health disorders using facial analysis.

How I built it

The pretrained-model was imported from Keras and re-trained on a dataset of 4900 photos of a diverse group of individuals featuring the 7 primary emotions. The application then compares an input photo (taken by the user) and uses the trained model to determine a probability distribution of emotions. The App was designed with Android Studio.

Challenges I ran into

  • We had issues increasing the accuracy of the model (it was stuck at 0.21% for a while)
  • We had difficulties with adding a camera api in the android application

Accomplishments that I'm proud of

  • Completing the application was a significant accomplishment
  • Getting the camera app to function

What I learned

  • Learned how to use a pre-trained model and how to resume training from a checkpoint
  • How to use the camera2 api for android development

What's next for MedAI

  • Cross-platform functionality
  • More biometrics incorporated into the system
  • Going beyond mental health (being able to detect physical disorders)

Note that the password for the website and its pages is 1234567890

Built With

Share this project:

Updates