Output Format
The two main projects that provide output files (that are not just videos) are FaceLandmarkImg and FeatureExtraction and their output is explained in detail in this page
The detected landmarks in an image are output in a text file, with a postfix _det_n if there are multiple faces in an image:
version: 1
npoints: 68
{
603.825 193.492
604.934 209.987
...
662.5 252.346
654.784 251.901
}
Each line corresponds to a facial landmark location in pixels (x y), the landmark indices follow the following scheme:
Single output file
Similarity aligned work
HOG features