This is an Autoencoder based neural network which converts rgb images into gray scale.
Configuration config.py contains all the paths and configuration
For Preprocessing preprocess_data.py : It takes in colored image as input and saves the resized colored image and respective gray scale image.
Training and Prediction train_and_test.py: Training is done using autoencoder, model will be stored and for testing images, gray scale image will be generated and stored in specified folder.
Folders: path : Location of training images color_path : Location of resized colored training images grap_path : Location of gray training images test_data_path : Location of testing images test_gray_path : Location of actual gray testing images test_output : Location of gray images generated by trained model model_folder : Location where model will get stored
Data: For training data is taken from : http://download.tensorflow.org/example_images/flower_photos.tgz It contains 3670 images of flowers, 5 folders for 5 types if flowers
For testing data is taken from : https://www.kaggle.com/olgabelitskaya/the-dataset-of-flower-images/data It contains one folder which has images of 210 flowers.
RUN: Install the requirements by running the setup file:
- sh setup.sh requirements.txt
- Change the name of folders in config.py
- sh rgb_to_gray.sh