We will use 7 emotions namely - We have use 7 emotions namely - 'Angry'đ , 'Disgust'đ¤ĸ, 'Fear'đą, 'Happy'đ, 'Neutral'đ, 'Sad'âšī¸, 'Surprise'đ˛ to train and test(validation) our algorithm using Convolution Neural Networks. You can directly use kaggle kernel notebook, so that you don't need to download the dataset. Also, you can download your code and run it in your local device directly.
Kaggle notebook - https://www.kaggle.com/mayank7900/notebook0a0917a828/edit
Tools and Libraries used -
- Install anaconda
- Jupyter notebook
- VSCode
- matplotlib
- numpy
- openCV
- keras
Given dataset of different expressions:
https://www.kaggle.com/jonathanoheix/face-expression-recognition-dataset
Screenshot of some angry expressions from the dataset on kaggle:
- source code -
mayankchaudhary26
/
Emotion_Detection_CNN_keras
Train and test our algorithm using Convolution Neural Networks and classify emotions in real-time.
Emotion_Detection_CNN_keras
We will use 7 emotions namely - 'Angry'đ , 'Disgust'đ¤ĸ, 'Fear'đą, 'Happy'đ, 'Neutral'đ, 'Sad'
âšī¸ , 'Surprise'đ˛ to train and test our algorithm using Convolution Neural Networks.
You need a dataset with different emotions :
Given dataset of different expressions on kaggle:
https://www.kaggle.com/jonathanoheix/face-expression-recognition-dataset
Emotion Detection â Classifying the emotion on the face as emotion_labels = ['Angry','Disgust','Fear','Happy','Neutral', 'Sad', 'Surprise']
Top comments (5)
Looks at bit off to me - on the last picture (from top left to bottom right) - I'd go with:
excited- definitely not fear,pained- sad is close I guess,pleading- again, sad is closesurprised- certainly not neutral,happy- spot on,confused- not fearHahaha right :) Well I have used this picture from the internet for the reference purpose only. My model will work fine with this dataset.
Poor training data set maybe? or perhaps the subtleties of the human face are beyond the model's ability to discern
Yo that's pretty neeaaat! Naaahhsss one!
Thanks ;)