import re

import sys

import csv

import os

import numpy as np

import matplotlib.pyplot as plt

from matplotlib.ticker import MultipleLocator

import pandas as pd

def draw_histo(df, col_name, x_start, bin_width, n_bins, title, output_dir, output_filename):

plt.get_figure().clf()

if __name__ == "__main__":

performance_csv = 'data/attack_example/performance_list.csv'

output_dir = '.'

with open (performance_csv) as csv_file:

df = pd.read_csv(csv_file,

dtype={

'img_num':str,

'epsilon':int,

'img_size': int,

'true_box_size':int,

'truth_iou_no_noise':float,

'truth_iou_noise_avg':float,

'ssim_avg':float})

print(df.head(10))

df = df[df['epsilon'] >= 0] # get rid of rows from failed attacks

eps_title = 'Epsilon Additive Noise Value to Beat HoG'

eps_filename = 'eps_hist.png'

draw_histo(df, 'epsilon', -16, 16, 16, eps_title, output_dir, eps_filename)

size_title = 'Image Size (Width x Height)'

size_filename = 'img_size_hist.png'

draw_histo(df, 'img_size', 400000, 100000, 21, size_title, output_dir, size_filename)

box_title = 'True Face Box Size (Width x Height)'

box_filename = 'true_box_size_hist.png'

draw_histo(df, 'true_box_size', -15000, 15000, 25, box_title, output_dir, box_filename)

iou_title = 'IoU between Found and True Box Prior to Attack'

iou_filename = 'truth_iou_no_noise.png'

draw_histo(df, 'truth_iou_no_noise', -0.05, 0.05, 21, iou_title, output_dir, iou_filename)

iou_title = 'IoU between Found and True Box After Noise Attack (Average)'

iou_filename = 'truth_iou_noise.png'

draw_histo(df, 'truth_iou_noise_avg', -0.05, 0.05, 21, iou_title, output_dir, iou_filename)

ssim_title = 'Structural Similarity (Average)'

draw_histo(df, 'ssim_avg', -0.05, 0.05, 21, ssim_title, output_dir, 'ssim.png')

bin_width = 0.01

x_start = -0.01

n_bins = 31

true_box_size_frac_plt = (df['true_box_size'] / df['img_size']).plot.hist(

bins=[bin_width * (x + 0.5) + x_start for x in np.arange(0, n_bins)],

edgecolor='black',

color='#88d498')

print('true_box_size_frac max: '

+ str((df['true_box_size'] / df['img_size']).max())

+ ', min: ' + str((df['true_box_size'] / df['img_size']).min()))

true_box_size_frac_plt.set_title('True Box Size as a Fraction of Total Image Size')

true_box_size_frac_plt.get_figure().savefig(output_dir + 'true_box_size_frac_hist.png')

true_box_size_frac_plt.get_figure().clf()