/hey-daug

A light weight data augmentation tool for training CNNs and Viola Jones detectors

Primary LanguagePythonMIT LicenseMIT

hey-daug

A light weight data augmentation tool for training CNNs and Viola Jones detectors (Haar Cascades). This tool inflates your data by up to six times. Use with care.

Steps for use:

  1. Set the parameters for images in constants.py

  2. Use data augmentation in your code:

import data_utils as daug
folders=[folder1, folder2] #list of folder paths where training images are saved, ex. ['./pos' , './neg']
daug.augment_and_save(folders)

If you would like to remove augmented images and keep originals, use:

du.remove_augmented_data(folders)