2017fall_smile

Intro

In this project, we aim to accomplish real time facial emotion recognition and possibly move to facial emotion generation.

refer to: real-time facial emotion recognition, face generation

Goal

1. Facial emotion recognition
   • for given face image, output its emotion label (happy, sad, angry, etc.)
2. Real time facial emotion recognition
   • use web-camera to recognize facial emotion in real time
3. Facial emotion generation (would be awesome to have this done)
   • for given plain face, generate face with wanted emotion

TimeLine

• September 20th, 2017

  • Download Opencv: tutorial

  • Learn how to use git and work with github

    please try to modify the following two lines:

    • zhenyu: find pytorch code
    • judy: find images

  • Use python package opencv to detect faces in a photo, something like

    

    for given image including faces, they should be exported as single face images

    the link might be useful: facial detection in 20 lines, feel free to use other tools

  • Find and download emotional faces datasets, something like

    

    Note the dataset you found should have uniform size, style and also labels regarding different emotions, the larger the dataset the better

    (please show some samples before downloading)

• September 27th, 2017

  • Understand convolutional neural network (CNN)

  • Use pytorch to train a CNN for hand-written digits classification

    the link might be useful: pytorch mnist example, feel free to use other sources

  • We will have joint meeting with Martin this week, we are suppose to show something that works.

• October 4th, 2017
  • Start dealing with emotional faces dataset process dataset into appropriate form for training purpose
  • Learn how to use polyps and run our classification model on server, probably use GPU
• October 11th, 2017
  • Have a workable facial emotions recognition (FER) model, means we should be able to recognize the emotion of given face image
  • Try deeper networks and larger dataset
  • We will have joint meeting with Martin this week
• October 18th, 2017

  • Extend the work to web-camera, we should be able to export images from real time web-camera and recognize the face emotions in the image

    A real time web-camera implementation by OpenCV Blog:

  • Do real time FER

• October 25th, 2017
  • Produce a visual demo for our real time facial emotions recognition, something like real-time facial emotion recognition
  • We will have joint meeting with Martin this week
• November 1st, 2017
  • Try other networks architecture, like resNet and denseNet
  • Improve our FER model accuracy level
• November 8th, 2017
  • Start work on facial emotion generation (FEG) model using GAN networks
  • Have a workable GAN demo from some existing projects
  • We will have joint meeting with Martin this week
• November 15th, 2017
  • Apply GAN networks to our FEG model
• November 22th, 2017
  • Produce a visual demo of FEG
  • We will have joint meeting with Martin this week
• November 29th, 2017
  • 😄

Dependency

• python basics/opencv/pytorch
• basic linear algebra
• neural networks

What you will get from this project

• Experience to work with git/github
• A real touch to deep learning/artificial intelligence
• A nice adding to you CV if your aim to a tech company