1 of 6

Facial Expression Recognition

Group 24

2 of 6

Introduction

  • Construct a facial expression recognition model
  • Recognize expression through webcam

3 of 6

Overview - Face Detection

Webcam Frame

Cv2 dnn model

(res10_300x300_ssd)

Some face images

(300x300)

Preprocessing

Face Expression Recognition Model

“happy”

“disgust”

4 of 6

Overview – training

FER2013�dataset

ResNet50(base model)

Fine tuning

Data augmentation

(flip, rotate, shift)

“happy”

5 of 6

Result

  • Reach 90% test accuracy

-> only reach at most 85% using ResNet as base model

  • Real time detection

-> work well using DNN model

6 of 6