Skip to content

ZhaoJ9014/Head-Pose-Estimator.Caffe

Repository files navigation

HeadPoseEstimator

  1. Introduction

The HeadPose Estimator predicts face landmarks (68 points) and head pose (3d pose, yaw, roll, and pitch).

  1. Install

caffe

dlib18.17

(cd your dlib folder

cd python_example

./compile_dlib_python_module.bat

add dlib.so to the PYTHONPATH)

opencv

numpy, pandas

  1. Usage

Unzip dlib-18.17.zip

Creat folder "img" to store your testing img

Creat folder "result" to store your visualized results

Creat folder "pose" to store your estimated poses (yaw angles) for each img

Pycharm run dlib-18.17/python_examples/landmarkPredict.py (modify the caffe_root path and testList.txt if needed. testList.txt is a file containing the path of the testing images.)

  1. Train

Modify train.prototxt and train_solver.prototxt files, train the model using your own data.