Head Pose Estimation using Transfer Learning

Head Pose Estimation is one of the most sought after problems in Computer Vision, as it is used in many real-time applications such as Advanced Driver Assistance System, Augmented Reality and many other Artificial Intelligence applications. But most of the existing systems are trained for controlled environment applications. They are not robust enough to apply them in real-time critical systems. We present a robust Head Pose Estimation System using convolution neural network augmented by transfer learning algorithm, which enables the system to adjust with the wild environment. Transfer learning approach also avoids the system from over-fitting and it also enables the model to be retrained for its respective application, in a fraction of time as compared to the time required to train an existing CNN model.