Results
Now that we have our verification and classification platform in place, now we will see how this will improve traditional methods currently available. Currently systems and sensors have a certain amount of false acceptance & false rejection which is factored into during the verification process. Plus we are using transfer learning rather than going for the traditional approach to reduce the training time.
Using tensor-flow we have achieved 94% accuracy on our model which can be seen in the graph below. Where the blue colored line shows the accuracy of the model on the 10% of the data input that were used as the testing set and the orange colored line is the accuracy of our model on the training set.