Implementation of the euler angles estimation via PFLD for faces cropped by the object detection model named yolov3.

0. Change cwd to the main folder of the project
1. Run python3 get_euler_angles.py with a source image:
   • --source https:// (you can give a web link to the image)
   • --source /path/to/file/image.jpg (single image file)
   • --source some/dir (directory with images)
   • default source will locate images in ./data/samples/
2. Watch command line to get the euler angles

To build your own docker container do as follows in your Unix terminal: 0. make build - this will only build the container

1. make start - this will build and start your newly built docker container
2. python3 get_euler_angles.py --source ./data/samples/driver.jpg - type it inside docker container terminal to see the cropped image and the corresponding euler angles
3. For a custom web image please refer to point 1 in Getting started guide
4. To see the results for a batch of images please also refer to point 1 in Getting started guide

Optionally, you can download a prebuilt docker container from Dockerhub via the link:

docker pull sharkzeeh/face:v1

and then do the same steps

pip3 install -r requirements.txt

Euler angles are calculated on the Wider Facial Landmarks in-the-wild (WFLW) face dataset. It contains 10000 faces (7500 for training and 2500 for testing) with 98 fully manual annotated landmarks.

PFLD: A Practical Facial Landmark Detector https://arxiv.org/pdf/1902.10859.pdf