This project was done as a part CS-574 (Computer Vision using Machine Learning) course at IIT Guwahati.
- Numpy
- Python 3.6
- Keras(with Tensorflow Backend)
- Pandas
- Scikit-Learn
We have tried out 4 types of Architecture namely :
- RandomForest
- Data was split into 23% test data and 77% training data.
- RMSE = 2.72725802
- Feed-forward Neural Network (with single hidden layer of 100 neuron)
- Convolutional Neural Network
- Convolutional Neural Network (with dropout on every layer)
Our result showed that CNN(with dropout) performed best on facial keypoint detection task followed by CNN(without dropout), RandomForest and Simple Feed Forward Network respectively.
- Abhinav Anshuman
- Aditya Gaurav
- Chirag Sodani
- Rahul Kumar