Skip to content

🔥🔥 Deep Learning Head Pose Estimation using PyTorch.

License

Notifications You must be signed in to change notification settings

MitchellX/deep-head-pose

 
 

Repository files navigation

headpose estimation



Hopenet is an accurate and easy to use head pose estimation network. Models have been trained on the 300W-LP dataset and have been tested on real data with good qualitative performance.

For details about the method and quantitative results please check the CVPR Workshop paper.



understand the euler angles

the output is: [yaw, pitch, roll]

yaw(red) ∈ [-180, 180]: represent face left & right rotation, left -, right +

pitch(green) ∈ [-90, 90]: represent head up & down move, up +, down -

roll(blue) ∈ [-180, 180]: represent head left & right shake, head close to left shoulder +, head close to right shoulder -



How to use

environment:python2.7 (I recommend virtualenv)

to use the virtualenv

pip install virtualenv

set up and activate a virtual environment

virtualenv headpose --python=2.7
source ./headpose/bin/activate

install the packages:

pip install -r requirentments_python27.txt

download the pretrained model

from the original author: (then put the pre-trained model in src folder)

300W-LP, alpha 1, robust to image quality

test on an image folders

python headpose_estimation_imageFolder.py -i ~/path/to/yours/

for more information:

python headpose_estimation_imageFolder.py -h

calculate the Euler-angle's L2 distance between two images:

you must change the source code in headpose_estimation_pairs_celeba_ffhq.py, here I give an example:

python headpose_estimation_pairs_celeba_ffhq.py celeba c3net  


Other pre-trained models:

(all the following messages are from the original Github repo)

300W-LP, alpha 1

300W-LP, alpha 2

300W-LP, alpha 1, robust to image quality

Acknowledgment

Thanks for the authors great work: https://github.com/natanielruiz/deep-head-pose

If you find Hopenet useful in your research please cite:

@InProceedings{Ruiz_2018_CVPR_Workshops,
author = {Ruiz, Nataniel and Chong, Eunji and Rehg, James M.},
title = {Fine-Grained Head Pose Estimation Without Keypoints},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2018}
}

Nataniel Ruiz, Eunji Chong, James M. Rehg

Georgia Institute of Technology

About

🔥🔥 Deep Learning Head Pose Estimation using PyTorch.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 98.9%
  • Shell 1.1%