Planing_path_algo/customDatasets.py at main · BKairat/Planing_path_algo · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
from __future__ import annotations

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import os
import shutil
from random import randint
from numpy.typing import ArrayLike
from random import uniform
from PIL import Image, ImageDraw
from tqdm import tqdm
import torch
from torch.utils.data import Dataset


def copy_file_to_folder(file: str, dest_folder: str) -> str:
    if not os.path.exists(dest_folder):
        os.makedirs(dest_folder)

    file_name = os.path.basename(file)

    dest_file = os.path.join(dest_folder, file_name)

    shutil.copy(file, dest_file)
    return dest_file


class DenoiseDataset(Dataset):
    def __init__(self,
                 file_csv: str = "out.csv",
                 root_dir: str = "./datasets/denoise_simple",
                 ):
        self.annotations = pd.read_csv(os.path.join(root_dir, file_csv))
        self.root_dir = root_dir

    def __len__(self):
        return len(self.annotations)

    def __getitem__(self, item: int | None) -> tuple:
        if item == None:
            item = randint(0, self.len()-1)
        noised_path = os.path.join(self.root_dir, self.annotations.iloc[item, 0])
        target_path = os.path.join(self.root_dir, self.annotations.iloc[item, 1])

        n_img = plt.imread(noised_path)
        t_img = plt.imread(target_path)
        return (n_img, t_img)

    def show(self, index: int = 0):
        fig, axis = plt.subplots(1, 2)
        noised, target = self[index]

        axis[0].imshow(noised)
        axis[0].set_title("noised path")

        axis[1].imshow(target)
        axis[1].set_title("target path")

        plt.show()

    def load_data(self) -> [[[], []], [[], []]]:
        train = [[], []]
        test = [[], []]
        len_ = len(self)
        print("loading train data:")
        for i in tqdm(range(round(0.8*len_))):
            train[0].append(plt.imread(os.path.join(self.root_dir, self.annotations.iloc[i, 0])))
            train[1].append(plt.imread(os.path.join(self.root_dir, self.annotations.iloc[i, 1])))
        print("complete!")
        print("loading test data:")
        for i in tqdm(range(round(0.8*len_), len_)):
            test[0].append(plt.imread(os.path.join(self.root_dir, self.annotations.iloc[i, 0])))
            test[1].append(plt.imread(os.path.join(self.root_dir, self.annotations.iloc[i, 1])))
        print("complete!")
        return train, test

class Generator():
    def __init__(self, root_dir: str = "./datasets/denoise_simple/",
                 size: int = 10000,
                 fname: str = "out",
                 augm: int = 0,
                 map_size: int = 64):
        self.annotations = os.path.join(root_dir, fname+".csv")
        self.root_dir = root_dir
        if not os.path.exists(self.root_dir):
            os.makedirs(self.root_dir)

        self.size = size
        self.augm = augm
        self.map_size = map_size

    @staticmethod
    def generate_random_points_circle(n: int, ref_point: ArrayLike, radius: float):
        angles = np.random.uniform(0, 2 * np.pi, n)
        x_coords = ref_point[0] + uniform(0, radius) * np.cos(angles)
        y_coords = ref_point[1] + uniform(0, radius) * np.sin(angles)
        return np.column_stack((x_coords, y_coords)).astype("int64")

    @staticmethod
    def generate_random_points_on_line(point1: np.array, point2: np.array, num_points: int) -> np.array:
        t_values = np.random.rand(num_points)
        line_points = point1 + t_values[:, np.newaxis] * (point2 - point1)
        return line_points

    @staticmethod
    def generate_random_points_line(point1: ArrayLike, point2: ArrayLike,
                                    distance_range: float = 0.05, n_points: int =10) -> list:
        x1, y1 = point1
        x2, y2 = point2

        if x2 == x1:
            x2 += 0.00001
        slope = (y2 - y1) / (x2 - x1)
        if slope > 2:
            slope = 2
        if slope < -2:
            slope = -2
        t_values = np.random.rand(n_points)
        rand_ps_on_line = point1 + t_values[:, np.newaxis] * (np.array(point2) - point1)

        noise = []
        for rand_p in rand_ps_on_line:
            new_x = rand_p[0] + np.sqrt(1 + slope ** 2) * (y2 - y1) * np.random.uniform(-distance_range, distance_range)
            new_y = rand_p[1] - np.sqrt(1 + slope ** 2) * (x2 - x1) * np.random.uniform(-distance_range, distance_range)
            noise.append([new_x, new_y])
        return noise

    def generate_noised_with_intersections(self, p_range: tuple = (3, 10)):
        data = {"noised": [], "target": []}
        print("generating dataset:")
        for i in tqdm(range(self.size)):
            image_target = Image.new("L", (self.map_size, self.map_size))
            noised = Image.new("L", (self.map_size, self.map_size))

            draw = ImageDraw.Draw(image_target)
            draw_noise = ImageDraw.Draw(noised)

            p = randint(p_range[0], p_range[1])

            a = (randint(0, 63), randint(0, 63))
            p = randint(3, 7)

            for i_ in range(p - 1):
                b = (randint(0, 63), randint(0, 63))
                draw.line([a, b], 255, width=0)
                rand_P = self.generate_random_points_line(a, b, distance_range=0.5, n_points=100)
                rand_P = np.array(rand_P)
                for p in rand_P:
                    p = tuple(p)
                    draw_noise.line([p, p], 255, width=0)
                a = b

            data["noised"].append(f"n_{i}.jpg")
            data["target"].append(f"t_{i}.jpg")

            image_target.save(os.path.join(self.root_dir, f"t_{i}.jpg"))
            noised.save(os.path.join(self.root_dir, f"n_{i}.jpg"))
        df = pd.DataFrame(data, index=None)
        df.to_csv(self.annotations, index=False)

    def generate_noised_simple(self, p_range: tuple = (7, 12)):
        data = {"noised": [], "target": []}
        print("generating dataset:")
        for i in tqdm(range(self.size)):
            image_target = Image.new("L", (self.map_size, self.map_size))
            noised = Image.new("L", (self.map_size, self.map_size))

            draw = ImageDraw.Draw(image_target)
            draw_noise = ImageDraw.Draw(noised)

            a = np.array((0, 0))
            p = randint(p_range[0], p_range[1])
            b = a
            for i_ in range(p - 1):
                while np.allclose(a, b):
                    b = np.array((randint(round(a[0]), 63), randint(round(a[1]), 63)))
                d = np.sqrt(sum((b - a) ** 2))
                d = d if float(d) > 1e-10 else 1e-8
                b = a + 10 * (b - a) / d
                draw.line([tuple(a), tuple(b)], 255, width=0)
                rand_P = self.generate_random_points_line(a, b, distance_range=0.5, n_points=200)
                rand_P = np.array(rand_P)
                for p in rand_P:
                    p = tuple(p)
                    draw_noise.line([p, p], 255, width=0)
                a = b
                if (a >= 63).any():
                    break

            data["noised"].append(f"n_{i}.jpg")
            data["target"].append(f"t_{i}.jpg")

            image_target.save(os.path.join(self.root_dir, f"t_{i}.jpg"))
            noised.save(os.path.join(self.root_dir, f"n_{i}.jpg"))
        df = pd.DataFrame(data, index=None)
        df.to_csv(self.annotations, index=False)