Illumina

Illumina is a Web App built to perform real time emotion detection on live camera feed. I have implemented a Convolutional Neural Network and trained the network on fer2013 dataset hosted by Kaggle. Detects upto 7 emotions for all identified faces in frame ['happy','neutral','fear','suprise','sad','disgust','anger']. Achieved 65%(top 10 percentile accuracy) on validation dataset . Then the saved model was used to perform emotion detection from webcamera input using OpenCV. Dataset Downloaded from : link

Directory Structure:

• templates : folder containing html files
• app.py : py file containing the server side code
• emotion_detection.py : neural network implementation
• predict.py : py file to run detection in a window instead of browser
• fer.h5 : saved model architecture
• fer.json : saved model weights

Steps to Run:

• Create a Virtual Environment and Install all dependencies from requirements.txt

  pip install -r requirements.txt

• Run app.py file

  python app.py

• Enter localhost address in browser.

Note : Make sure there is good lighting in the room to get better accuracy.

Thats it ^_^

Demo