- Qi Hu : qihucn@uw.edu
- Xinyu Zhao : xinyu94@uw.edu
- Cunzhi Ren : renc@uw.edu
- This is a project for face detection and facial expression classification
- We use Haar Cascade algorithm provided by openCV to detect faces in the video/image
- We compare different model for facial expression classification: kNN, SVM and CNN.
- A demo is provided, where you can run the detection and classification in real time, and the program will replace the face with a corresponding emoji sign.
- For more details, please refer to the POSTER 'poster_facedetect.pdf'
-
Train (CNN model) : run script cnn.py/cnn_keras.py to train a CNN classification model (cnn.py is the tensorflow version, cnn_keras.py is the keras version)
-
Test (CNN model): the model will be ran on test set each epoch. to draw the curve of training and testing process, run eval.py
-
kNN, SVM : knn.py is for kNN testing, svm.py is for SVM training and testing;
-
Demo : run the script demo.py to start the demo. This demo will use a camera to obtain real time video, and automatically replace a detected face with corresponding emoji sign
-
(model parameters can be adjusted in scripts cnn.py/cnn_keras.py, knn.py and svm.py)