Human-facial emotion analysis and recognition with CNN

This project will detect 7 types of facial emotion, angry, disgust, fear, happy, neutral, sad, surprise.

Packages that I have used and need to be installed

• pip install numpy
• pip install opencv-python
• pip install keras
• pip3 install tensorflow
• pip install pillow
• I have also used HaarCascade to make boundary around the face.

The FER2013 dataset can be downloaded from the link below

• from the below link, the dataset needs to be put in the data folder under the project directory
• https://www.kaggle.com/msambare/fer2013

Train Emotion detector

• used face expression images in the FER2013 Dataset
• Run the command --> python TrainEmotionDetector.py ; to train the ML Model

It took around 1hour 45minutes for my laptop with Ryzen 7 processor, 16gb RAM and Nvidia GeForce GTX 1650 GPU. After Training, we will find the trained model structure and weights are stored in the project directory as:- emotion_model.json emotion_model.h5

Now we can copy these two files created, and paste in model folder in the project directory.

Run your emotion detection test file

Run the command --> python TestEmotionDetector.py ; to see the output. I have made 2 options to view the result(which can be seen in TestEmotionDetector.py file,

• cap = cv2.VideoCapture(0) ; this can be used for Human-facial emotion prediction using laptop/pc camera in real time.

• cap = cv2.VideoCapture("{Location of the video}") ; this can be used for Human-facial emotion prediction using a video from computer's memory. Just the link to the video needs to be pasted in the code.

  Link to the video used: https://drive.google.com/file/d/1XmfWlPhj4-2wyRofiqAO_kqNPiOUmVSW/view?usp=sharing ; Link of the output video created: https://drive.google.com/file/d/135Vhj6re6WcUQpGEm190A5FREJ4UWqRs/view?usp=sharing ;