Inhalt der func.py

import pandas as pd
import numpy as np
import os
from tensorflow.keras.preprocessing.image import load_img, img_to_array
import matplotlib.pyplot as plt
import cv2
from shapely.geometry import Polygon, box
import shutil

def get_list_imag_label_files(label_path):
    ''' 
    input: the path to the label directory 
    return: *tuples*: list of labels and list of images with 'helicopter' and 'plane'
    description: reduce file by 'helicopter' and 'plane' only
    ''' 
    img_list = list()
    text_list = list()
    label_list = os.listdir(label_path)
    for filename in label_list:
        path =  label_path + "\\" + filename
        with open(path) as f:
            if 'plane' in f.read():
                png = str.split(filename, ".")[0] + ".png"
                txt = str.split(filename, ".")[0] + ".txt"
                text_list.append(txt)
                img_list.append(png)
            elif 'helicopter' in f.read():
                png = str.split(filename, ".")[0] + ".png"
                txt = str.split(filename, ".")[0] + ".txt"
                text_list.append(txt)
                img_list.append(png)
    return img_list, text_list

# FUNCTION: get_list_boxes_labels
def get_list_boxes_labels(label_path, label_file):
    ''' 
    input: the path to the label directory and the label file name 
    return: list of bounding box and list of related label
    description: open label files, extract bounding box coordinates and translate label into 0,1
    '''
    boxes = list()
    labels = list()
    path = label_path + "/" + label_file
    with open(path) as f:
        for line in f:
            if 'plane' in line:
                box = str.split(line)[0:8]
                boxes.append(box)
                #for plane label = 0
                labels.append(0)
            elif 'helicopter' in line:
                box = str.split(line)[0:8]
                boxes.append(box)
                #for helicotper label = 1
                labels.append(1)
    return boxes, labels


def get_regular_yolov(myArray,np_img,rect_plot=False):
    '''
    takes numpy array of shape (4,2) from the label-files and the image as array
    returns yolov5 format with normalized centriod, width and height

    if rect_plot is False a yolo format is returned
    if rect_plot is True a dictionary with coordinates for ploting is returned

    first calculation of mean of two lowest and two heighest values for x and y axis
    second calculation of centroid coordinates with shapely function Polygon
    third extraction of picture shape from image (as array)
    fourth normalization according to yolo requirements for labeling
    '''

    x_max = np.mean(np.sort(myArray[:,0])[-2:])
    x_min = np.mean(np.sort(myArray[:,0])[:2])
    y_max = np.mean(np.sort(myArray[:,1])[-2:])
    y_min = np.mean(np.sort(myArray[:,1])[:2])
    centroid = Polygon(myArray).centroid.coords[0]

    w_img = np_img.shape[1]
    h_img = np_img.shape[0]

    norm_centroid_x = centroid[0] / w_img
    norm_centroid_y = centroid[1] / h_img
    norm_width = abs(x_max-x_min) / w_img
    norm_height = abs(y_max-y_min) / h_img

    if rect_plot:
        return {
            'centroid':(np.rint(centroid[0]).astype(int),np.rint(centroid[1]).astype(int)),
            'coord_one':(np.rint(x_max).astype(int),np.rint(y_max).astype(int)),
            'coord_two':(np.rint(x_min).astype(int),np.rint(y_min).astype(int))
            }
    else:
        return (norm_centroid_x,norm_centroid_y,norm_width,norm_height)