/less-simple-fer-dl

Play stuff now advanced to neural networks along with more complex sampling methods to solve facial expression recognition.

Primary LanguageJupyter Notebook

less-simple-fer-dl

Project summarization

Update from the Basic Machine Learning project

  • Apply DL in solving Facial Expression Recognition problem
  • Models used: VGGFace, VGGFace2
  • Data augmentation methods: Random Oversampling, ADASYN, SMOTE
  • Datasets: FER2013, FER+
  • Create a simple web app using Flask simulating a virtual assistant which will response with audio to the corresponsing expression

Pipeline

image

More details

Read our thesis here

How to reproduce the demo

Make sure webcam, speaker is available
Create the conda environment: conda myenv create -f environment.yml
Navigate to voice-response/demo/web/flask
Run python thread_demo.py

Environments

Windows, Conda, Python