Skip to content

Latest commit

 

History

History
79 lines (42 loc) · 3.06 KB

README.md

File metadata and controls

79 lines (42 loc) · 3.06 KB
Raw

BU EC601-Final-Project

Project Description: an Andorid and PC App that translates user's expression into emojis.

Our application can detect 6 basic human expressions: happy, sad, surprise, neutral, fear, angry.

Project traits:

  1. PC real-time expression detection

  2. multiple faces expression detection both on PC and Android

  3. image expression detection on Android

  4. multiple expressions emoji matching (combination emojis like happy-surprise, angry-surprse) or sigle(happy) expression matching

  5. our model architecture:

gray-scale 48x48 image (cropped) >> 3 sequential conv.layers (32 feat. maps per layer) >> maxpool >> 3 sequential conv.layers (64 feat. maps per layer) >> maxpool >> 3 sequential conv.layers (128 feat. maps per layer) >>maxpool >> 2 sequential dense layers (20% dropout) >> softmax output

Authors:

Minghe Ren ([email protected])

Simin Zhai ([email protected])

Tianhen Hu ([email protected])

Xueying Pan ([email protected])

Directory Descriptions:

emojis : pngs of emojis (we're updating more)

model: training models and opencv models (Models with various combinations were trained and evaluated using GPU computing g2.2xlarge on AWS)

datasets: training datasets and codes we used

emojime_app -- android application

File descriptions:

  1. The haar-cascade_frontalface_default.xml in OpenCV contains pre-trained filters and uses Adaboost to quickly find and crop the face.

  2. real-time.py allows you to run real-time face and expression detection and save the predection values into emotion.txt

  3. live-plotting.py allows you to draw data saved in emotion.txt

  4. emojis.py -- This is our core code which finishes the expression detection and emojis matching.

  5. model.h5 and model.json -- our trained model and all kinds of model parameters like weights

  6. my_model.pb -- for android implantation

  7. Nariz.xml -- open source 25x15 Nose detector computed with 7000 positive samples

  8. feedback of our colleagues -- the feedback summary of our colleagues

  9. Emoji App Test based on Android.pdf and testemoji.xlsx are our test files

10.app-debug.apk -- our android test app

11.Map user's expressions into emojis.pdf -- our poster

How to run this project:

In this directory, try:

python real-time.py haar-cascade_frontalface_default.xml

python emojis.py haar-cascade_frontalface_default.xml

References:

  1. "Dataset: Facial Emotion Recognition (FER2013)" ICML 2013 Workshop in Challenges in Representation Learning, June 21 in Atlanta, GA.

  2. "Andrej Karpathy's Convolutional Neural Networks (CNNs / ConvNets)" Convolutional Neural Networks for Visual Recognition (CS231n), Stanford University.

  3. Srivastava et al., 2014. "Dropout: A Simple Way to Prevent Neural Networks from Overfitting", Journal of Machine Learning Research, 15:1929-1958.

  4. Duncan, D., Shine, G., English, C., 2016. "Report: Facial Emotion Recognition in Real-time" Convolutional Neural Networks for Visual Recognition (CS231n), Stanford University.

  5. Jostine Ho, "mememoji", Interactive Web App Mememoji is an interactive emotion recognition system that detects emotions based on facial expressions.