Head-Pose-Estimation

In this project I've worked on AFLW2000-3D dataset which is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks, with various head poses for humans and animations. By using mediapipe to extract faces landmarks which are 468 points in 3D, but we used x-axis & y-axis. The regression model of predicting the 3 angles (pitch - yaw - roll) of head pose estimation was XGboost Regressor.

Data

You can download from http://www.cbsr.ia.ac.cn/users/xiangyuzhu/projects/3DDFA/Database/AFLW2000-3D.zip This data contains different pictures for human faces with different poses.

Steps:

1. Preparing data for model training ( we used MediaPipe and CV2 libraries for extracting points and for face detection from pictures).
2. Spliting the data to training, validation and testing.
3. Using regression model.
4. Detecting a random pic to validate the model from dataset.

Conc:

Humans:

• https://drive.google.com/file/d/1RwTD1i-DmeRcXFli0syQhl9FwazVI30f/view?usp=sharing

• https://drive.google.com/file/d/1R0c40IpKj0w7qftg1swEiE-hb-PUg2qW/view?usp=sharing

Animation:

• https://drive.google.com/file/d/1XTgzZG1BckDbh1RiaLAkxQ6lxp183B-m/view?usp=sharing

For more information about mediapipe:

https://google.github.io/mediapipe/solutions/face_mesh.html