fazeVaib / gesture-digit-recog

Tracks an object and records its trajectory to predict which shape is drawn by that object using SVM.

gesture-digit-recog

Recognize the digits in real-time through shapes made in front of camera

Contents

• "model_train.py" consists of the model trained using MNIST dataset implemented using SVM.
• "mnist_clf_svm.pkl" file consists of the saved model, which is later being loaded into the prediction function for preeiction.
• "camera_pred.py" is the file that carries out the prediction by accessing the camera. It displays the result in the original frame itself.

Requirements

• openCV
• Numpy
• Scikit-image
• Sklearn
• Tensorflow
• Keras

Operation

• The model is pre-trained and saved, but you can also train it with your own data.
  
• Run file 'camera_pred.py' in terminal.

* Use a green object to draw the digits in front of camera.
* press 'C' to clear the input and 'Q' to quit the program.