protosc

Python library for automatic feature extraction. It is currently under active development. The overall pipeline looks as follows:

• preprocessing of the data
• feature extraction
• feature selection

Below we list from all these categories, what is currently developed or under development.

IO

• Read image from file.

Preprocessing

Currently only image processing preprocessing steps are included.

• convert to grey scale
• Apply Viola-Jones algorithm for face detection
• Cut a circle in the middle (face after VJ)

Feature extraction

• Fourier transformation of image

Feature selection

• Filter feature selection
• Combination of results from filter feature selection to final feature selection.

Pipeline

The pipeline system allows for easy definition of pipelines:

pipe1 = ReadImage()*GreyScale()*ViolaJones(add_perc=15)*FourierFeatures()
pipe2 = ReadImage()*ViolaJones()*FourierFeatures()
pipe_complex = pipe1 + pipe2
features = pipe_complex.execute(x)  # Here x can be a list of filenames