This project is a C++ reimplementation of face alignment in 3000fps in the CVPR 2014 paper: Face Alignment at 3000 FPS via Regressing Local Binary Features. .
Update openMP support !!!
I modify my code to support openMP. You can use it in GCC(Linux) or in VS (Windows).
If you use it in Linux, you should comment or uncomment
FIND_PACKAGE( OpenMP REQUIRED) in CmakeLists.txt.
If you use it in Windows, you can directly use it.
I add a VS project.
Download datasets and get Path_Images.txt as jwyang/face-alignment.
To compiler the program: go to folder
To train a new model: set global parameters, model path, train database name in
To test a model on dataset: set model path, test dataset name in
How to get the bounding box of image ?
I use the face detector in OpenCV to get the bounding box.You can use any detector to get the bounding box but you must provide a bounding box of similar measure with the training data.
How about the liblinear?
I add the liblinear source code as the project code. So you can directly compiler this project and don't need to consider to compiler this library.
If you have any question, you can create an
issue on GitHub.
Or you can email email@example.com