A utility to load facial landmark dataset from a single file

[imagesPaths, landmarks, success] = cv.Facemark.loadTrainingData2(filename)
[...] = cv.Facemark.loadTrainingData2(..., 'OptionName',optionValue, ...)

Input

• filename The filename of a file that contains the dataset information. Each line contains the filename of an image followed by pairs of x and y values of facial landmarks points separated by a space.

Output

• imagesPaths A cell-aray where each element represent the filename of image in the dataset. Images are not loaded by default to save memory.
• landmarks The loaded landmark points for all training data {{[x,y], ...}, ...}.
• success success flag.

Options

• Delim Delimiter between each element, the default value is a whitespace ' '.
• Offset An offset value to adjust the loaded points. default 0.0

This dataset format simply consists of a single file, where each line contains an image path followed by list of x/y coordinates of its corresponding annotation. Number of points of each sample is not included, and the code will load all value pairs until the end of line.

Example of file format expected:

/path/to/image1.jpg 336.820955 240.864510 334.238298 260.922709 ...
/path/to/image2.jpg 376.158428 230.845712 376.736984 254.924635 ...
...