helper.py

""" Includes helper functions that are used to pre-process observations, or
    to save figures, lists etc. """

import numpy as np

import matplotlib

#matplotlib.use('Agg') #hopw this is gonna work

import matplotlib.pyplot as plt

from skimage.transform import resize
from skimage.color import rgb2gray

import tensorflow as tf

def setup_saver(W, b):
    features = W.copy()
    features.update(b)
    saver = tf.train.Saver(features)

    return saver


#Helper functions
def save_figs(rewardList, stepsList, parameter, i):
    """ Creates matplotlib figures out of the accumulated lists """

    plt.interactive(False)
    plt.plot(rewardList)
    if parameter['SAVE_FIGS']:
        plt.savefig( 'rewards{:d}.png'.format(i) )
    else:
        plt.ylabel('rewards')
        plt.show()
    
    plt.plot(stepsList)
    if parameter['SAVE_FIGS']:
        plt.savefig( 'steps{:d}.png'.format(i) )
    else:
        plt.ylabel('steps')
        plt.show()

def preprocess_image(observation):  #takes about 20% of the running time!!!
    """ Grayscale, downscale and crop image for less data wrangling """
    #consider transfering this to TF
    out = rgb2gray(observation)    #takes about 5% of the running time!!!               #2s
    out = resize(out, (110, 84))    #takes about 9% of running time!!!
    return out[13:110 - 13, :]