123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191 |
- #!/usr/bin/env python3
- # -*- coding: utf-8 -*-
- """
- Created on Fri Sep 14 21:02:42 2018
- @author: jbuisine
- """
- from __future__ import print_function
- import sys, os, getopt
- import numpy as np
- import random
- import time
- import json
- from PIL import Image
- from ipfml import processing
- from ipfml import metrics
- from skimage import color
- import matplotlib.pyplot as plt
- from modules.utils import config as cfg
- config_filename = cfg.config_filename
- zone_folder = cfg.zone_folder
- min_max_filename = cfg.min_max_filename_extension
- # define all scenes values
- scenes_list = cfg.scenes_names
- scenes_indexes = cfg.scenes_indices
- choices = cfg.normalization_choices
- path = cfg.dataset_path
- zones = cfg.zones_indices
- seuil_expe_filename = cfg.seuil_expe_filename
- metric_choices = cfg.metric_choices_labels
- max_nb_bits = 8
- def display_data_scenes(nb_bits, p_scene):
- """
- @brief Method display shifted values for specific scene
- @param nb_bits, number of bits expected
- @param p_scene, scene we want to show values
- @return nothing
- """
- scenes = os.listdir(path)
- # remove min max file from scenes folder
- scenes = [s for s in scenes if min_max_filename not in s]
- # go ahead each scenes
- for id_scene, folder_scene in enumerate(scenes):
- if p_scene == folder_scene:
- print(folder_scene)
- scene_path = os.path.join(path, folder_scene)
- config_file_path = os.path.join(scene_path, config_filename)
- with open(config_file_path, "r") as config_file:
- last_image_name = config_file.readline().strip()
- prefix_image_name = config_file.readline().strip()
- start_index_image = config_file.readline().strip()
- end_index_image = config_file.readline().strip()
- step_counter = int(config_file.readline().strip())
- # construct each zones folder name
- zones_folder = []
- # get zones list info
- for index in zones:
- index_str = str(index)
- if len(index_str) < 2:
- index_str = "0" + index_str
- current_zone = "zone"+index_str
- zones_folder.append(current_zone)
- zones_images_data = []
- threshold_info = []
- for id_zone, zone_folder in enumerate(zones_folder):
- zone_path = os.path.join(scene_path, zone_folder)
- current_counter_index = int(start_index_image)
- end_counter_index = int(end_index_image)
- # get threshold information
- path_seuil = os.path.join(zone_path, seuil_expe_filename)
- # open treshold path and get this information
- with open(path_seuil, "r") as seuil_file:
- seuil_learned = int(seuil_file.readline().strip())
- threshold_info.append(seuil_learned)
- # compute mean threshold values
- mean_threshold = sum(threshold_info) / float(len(threshold_info))
- print(mean_threshold, "mean threshold found")
- threshold_image_found = False
- # find appropriate mean threshold picture
- while(current_counter_index <= end_counter_index and not threshold_image_found):
- if mean_threshold < int(current_counter_index):
- current_counter_index_str = str(current_counter_index)
- while len(start_index_image) > len(current_counter_index_str):
- current_counter_index_str = "0" + current_counter_index_str
- threshold_image_found = True
- threshold_image_zone = current_counter_index_str
- current_counter_index += step_counter
- # all indexes of picture to plot
- images_indexes = [start_index_image, threshold_image_zone, end_index_image]
- images_data = []
- print(images_indexes)
- low_bits_svd_values = []
- for i in range(0, max_nb_bits - nb_bits + 1):
- low_bits_svd_values.append([])
- for index in images_indexes:
- img_path = os.path.join(scene_path, prefix_image_name + index + ".png")
- current_img = Image.open(img_path)
- block_used = np.array(current_img)
- low_bits_block = processing.rgb_to_LAB_L_bits(block_used, (i + 1, i + nb_bits + 1))
- low_bits_svd = metrics.get_SVD_s(low_bits_block)
- low_bits_svd = [b / low_bits_svd[0] for b in low_bits_svd]
- low_bits_svd_values[i].append(low_bits_svd)
- fig=plt.figure(figsize=(8, 8))
- fig.suptitle("Lab SVD " + str(nb_bits) + " bits values shifted for " + p_scene + " scene", fontsize=20)
- for id, data in enumerate(low_bits_svd_values):
- fig.add_subplot(3, 3, (id + 1))
- plt.plot(data[0], label='Noisy_' + start_index_image)
- plt.plot(data[1], label='Threshold_' + threshold_image_zone)
- plt.plot(data[2], label='Reference_' + end_index_image)
- plt.ylabel('Lab SVD ' + str(nb_bits) + ' bits values shifted by ' + str(id), fontsize=14)
- plt.xlabel('Vector features', fontsize=16)
- plt.legend(bbox_to_anchor=(0.5, 1), loc=2, borderaxespad=0.2, fontsize=14)
- plt.ylim(0, 0.1)
- plt.show()
- def main():
- if len(sys.argv) <= 1:
- print('Run with default parameters...')
- print('python generate_all_data.py --bits 3 --scene A')
- sys.exit(2)
- try:
- opts, args = getopt.getopt(sys.argv[1:], "hb:s", ["help=", "bits=", "scene="])
- except getopt.GetoptError:
- # print help information and exit:
- print('python generate_all_data.py --bits 4 --scene A')
- sys.exit(2)
- for o, a in opts:
- if o == "-h":
- print('python generate_all_data.py --bits 4 --scene A')
- sys.exit()
- elif o in ("-b", "--bits"):
- p_bits = int(a)
- elif o in ("-s", "--scene"):
- p_scene = a
- if p_scene not in scenes_indexes:
- assert False, "Invalid metric choice"
- else:
- p_scene = scenes_list[scenes_indexes.index(p_scene)]
- else:
- assert False, "unhandled option"
- display_data_scenes(p_bits, p_scene)
- if __name__== "__main__":
- main()
|