Dataset.py

import os
import random

from joblib import Parallel, delayed
import numpy as np
import glob
import cv2
import csv


# ------------ Data Initialization --------------
# Dataset class for easy access


class Dataset:
    def __init__(self, img, filename, filepath, descripList, feature):
        self.img = img
        self.filename = filename
        self.filepath = filepath
        self.descripList = descripList
        self.feature = feature

    def __repr__(self):
        return "Image: % s, Description List: % s, Feature: %s" % (self.filename, self.descripList, self.feature)


# Image Description class for easy access


class ImageDescrip:
    def __init__(self, descrip, imgList):
        self.descrip = descrip
        self.imgList = imgList

    def __repr__(self):
        return "Image Description: % s, ImageList: % s" % (self.descrip, self.imgList)


# Create Grayscale


def convert_greyscale(img):
    '''Convert image to Grayscale\n
       Parameters
           img: image object
    '''
    # we have to convert it to grayscale
    if (len(img.shape) > 2):
        img = img.astype(float)
        # RGB to grayscale convertion using Luminance
        # We are extracting R,G & B channels from patch and
        # using grayscale conversion algorithm that is used in OpenCV’s cvtColor().
        img = img[:, :, 0] * 0.3 + img[:, :, 1] * 0.59 + img[:, :, 2] * 0.11

    # converting to uint8 type for 256 graylevels
    img = img.astype(np.uint8)

    # normalize values can also help improving description
    i_min = np.min(img)
    i_max = np.max(img)
    if (i_max - i_min != 0):
        img = (img - i_min) / (i_max - i_min)

    return img


# Create DataSet Object


def create_dataset_object(img_file):
    '''Creates object of Class: DataSet and invokes opencv image read\n
       Parameters
           file: path for an image
    '''
    img = cv2.imread(img_file)
    return Dataset(img, img_file[-10:-4], img_file, [], None)


def create_random_dataset_object(img_file):
    '''Creates object of Class: DataSet and uses random values for the image\n
       Parameters
           file: path for an image
    '''
    img = np.random.randint(low=0, high=255, size=(375, 500, 3), dtype=np.uint8)
    return Dataset(img, img_file, img_file, [], None)


# Create Image Data Set


def create_dataset(img_folder_path):
    '''Import Images from file location and create DataSet class array: image_dataset
    '''
    # Image Data Set location setting
    imdir = os.getcwd() + img_folder_path + '*.jpg'

    # Reading images from location to a class array
    image_dataset = Parallel(n_jobs=-1)(delayed(create_dataset_object)(file)
                                        for file in glob.glob(imdir))

    return image_dataset, len(image_dataset)


def create_random_dataset():
    '''Create random images and create DataSet class array: image_dataset
    '''
    # Reading images from location to a class array
    image_dataset = Parallel(n_jobs=-1)(delayed(create_random_dataset_object)(str(i))
                                        for i in range(5011))

    return image_dataset, len(image_dataset)


# Extract ImageList


def extract_imageDescrip(img_class_set_names, img_class_path):
    '''Extracting Image List for different Image Descriptions.
       Import csv files from file location and create ImageDescrip class, 
       array: image_classSet
    '''
    image_classSet = []
    temp_img_classSet = []
    classSet = []
    for descrip in img_class_set_names:
        entry = []
        path_text = os.getcwd() + img_class_path + \
                    descrip["name"] + '_trainval.csv'
        with open(path_text, 'r') as file:
            my_reader = csv.reader(file, delimiter=',')
            for row in my_reader:
                if (row[1] == '-1'):
                    continue
                if (row[0] == '000005'):
                    row[0] = '000005'
                entry.append(row[0])

        image_classSet.append(ImageDescrip(descrip["name"], set((entry))))
    #     if descrip["type"] not in classSet:
    #         classSet.append(descrip["type"])
    #     temp_img_classSet.append(ImageDescrip(descrip["type"], entry))

    # for classType in classSet:
    #     entry = set()
    #     for imgdescrip in temp_img_classSet:
    #         if classType == imgdescrip.descrip:
    #             entry.update(imgdescrip.imgList)
    #     image_classSet.append(ImageDescrip(classType, set((entry))))

    return image_classSet


def extract_random_image_descrip(img_class_set_names):
    '''Assigning random images to random classes
    '''
    image_classSet = []
    num_of_classes = len(img_class_set_names)
    temp_img_classSet = [[] for i in range(num_of_classes)]

    for i in range(5011):
        temp_img_classSet[random.randint(0, num_of_classes - 1)].append(str(i))

    temp = 0
    for descrip in img_class_set_names:
        image_classSet.append(ImageDescrip(descrip["name"], temp_img_classSet[temp]))
        temp += 1

    return image_classSet