1)This project is to estimate human head pose in video.
2)The dataset used here is HPEG(A Natural Head Pose abd Eye Gaze Dataset).This dataset consists of 2 sessions. I have used session a. Session b also includes Gaze data.
3)In Session a,each video last for almost 6 seconds. There is also a "data"file for each video,which consists of head angle(Yaw).
4) I have used Probabilistic Graphical Model for classification. In more datails,I have used Conditional Random Fields in terms of HCRF Library.
5)For this phase I have used the following paper:"Head Pose Estimation Using Histogram of SIFT Descriptors ". So I have used SIFT Features too.
6) the pipeline is as follows:
- Video Framing
- Face extraction
- feature extraction : SIFT and then BOW(Bag of Words)
- Classification using CRF