Human Classifier

Convolutional neural network that classifies human characteristics by the image of its face.

Count, nature and classes of categories determines by you.

make_pickles.py - script that makes pickles from images.

convolutional_neural_network.py - script that builds and trains convolutional neural network models.

validator.py - script that uses prediction of some models (builded and trained by using convolutional_neural_network.py) and renames every image in folder by prediction (e.g. 10_male_adult_blonde).

utkface_dataset_parser.py - script for parsing UTKFace dataset .

Used libraries:

For correct work you must install your directories structure following this scheme:

where:

You must create each colorful field manually, but light grey fields are optional: scripts may create them for you.

Gradient fields in some cases would be created by scripts, but other cases you must create it by your own (which cases exactly determines by the first tree (light gray - manually, colorful - opposite)).

Documents on russian:

Terms of reference (obsolete)

Research report (obsolete)