https://github.com/shashikg/emotionrecognition
Detects Real Time Human Facial Emotions Trained on ICML 2013 dataset of Facial Expression Recognition Challenge on kaggle.
https://github.com/shashikg/emotionrecognition
cnn-keras convolutional-neural-networks emotion-detection emotion-recognition facial-expression-recognition machine-learning opencv
Last synced: about 2 months ago
JSON representation
Detects Real Time Human Facial Emotions Trained on ICML 2013 dataset of Facial Expression Recognition Challenge on kaggle.
- Host: GitHub
- URL: https://github.com/shashikg/emotionrecognition
- Owner: shashikg
- Created: 2018-12-07T06:18:48.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-06-13T20:12:55.000Z (almost 5 years ago)
- Last Synced: 2025-03-27T04:33:26.828Z (2 months ago)
- Topics: cnn-keras, convolutional-neural-networks, emotion-detection, emotion-recognition, facial-expression-recognition, machine-learning, opencv
- Language: Jupyter Notebook
- Homepage:
- Size: 60.3 MB
- Stars: 5
- Watchers: 0
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# Real Time Human Emotion Recognition
This repo contains files related to my project on emotion recognition carried during the end of my 5th semester as a hobby project. Presently, its capable of extracting faces from a web cam stream and classify them into 7 different moods i.e. Angry, Disgust, Fear, Happy, Sad, Surprise and Neutral. The face detection module uses already trained Haar-Cascade/ dnn based face detector from OpenCV. And Classifier was trained on the **ICML 2013** dataset of [Facial Expression Recognition Challenge](https://www.kaggle.com/c/challenges-in-representation-learning-facial-expression-recognition-challenge) on kaggle.
**Link to Demo Video:** [https://youtu.be/XVQSMbeBGZQ](https://youtu.be/XVQSMbeBGZQ)
## Training Results
**Accuracy Achieved on Test Data:** 65.34 %
## Proposed Neural Network Architecture
## Usage
Download the **Dataset** from [here](https://www.kaggle.com/c/challenges-in-representation-learning-facial-expression-recognition-challenge) and put **fer2013.csv** inside the data folder. Then install all the dependencies in a virtual environment
```
$ virtualenv --system-site-packages -p python3 ./venv
$ source ./venv/bin/activate
$ pip install -r requirements.txt
```
Then run the following commands to test the module
```
# To train a new model
$ python emotion_trainer.py
# To run the emotion recognition module
$ python emotion_recogniser.py
```
### Selecting face-detector
In **emotion_recogniser.py** change the following line as per your choice:
```
# for dnn based use this
faces = get_faces(img, method='dnn')
```
```
# for haar-cascade based use this
faces = get_faces(img, method='haar')
```
## Directory Structure
```
├── saved_model
│ └── cnn0.h5 ..................................:: Trained CNN model
│ └── cnn1.h5 ..................................:: Trained CNN model on mirror images
│ └── ensemble.h5 ..............................:: Trained NN Model on prediction from above two
│ └── haarcascade_frontalface_default.xml ......:: OpenCV Haar-Cascade model for frontfaces
│ └── opencv_face_detector_uint8.pb ............:: OpenCV trained model for faces
│ └── haarcascade_frontalface_default.xml ......:: OpenCV trained model for faces
├── detectfaces.py ...............................:: Contains function to call face detector and
│ crop & resize the detected faces to 48x48 size
├── emotion_trainer(.py/ .ipynb) .................:: This will train the designed Neural Network model on the given dataset
├── emotion_recogniser.py ........................:: This is the main program which will show real-time emotion classification
│ using the models saved in the directory 'saved_model/'
├── plot_confusion_matrix.py .....................:: Used to generate the confusion matrix of the trained model
├── img/ .........................................:: Contains some images related to results and for testing
├── requirements.txt
└── README.md
```