ForestFireSegmentation/utils.py at main · PixelWeaver/ForestFireSegmentation · 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
import functools
import json
import math
import os
import cv2
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd

def get_venn_circle_height(num):
    baseHeight = 20
    unit_a = math.pow(baseHeight / 2, 2) * math.pi / 42
    output_a = unit_a * (42 + math.pow(num - 42, 1.1))
    output_r = math.sqrt(output_a / math.pi)
    return output_r * 2

def paths_from_name(name):
    return (
        f"dataset/img/{name}.png",
        f"dataset/gt/{name}.png",
        f"dataset/nir/{name}.png"
    )

def load_row(row):
        rgb_path, gt_path, nir_path = paths_from_name(row[1])

        nir_im = None
        if nir_path is not None:
            nir_im = cv2.imread(nir_path)

        return (cv2.imread(rgb_path), np.expand_dims(cv2.imread(gt_path, flags=cv2.IMREAD_GRAYSCALE), axis=2)/255, nir_im)

def ensure_folders_exist(paths):
    for path in paths:
        if not os.path.isdir(path):
            os.mkdir(path)

def plot_history(name, plot_val=True):
    # Ensure output folder existence
    ensure_folders_exist([
        "figures",
        f"figures/{name}"
    ])

    # Parse and plot all metrics
    with open(f'histories/{name}.json', 'r') as fp:
            data = json.load(fp)

            df = pd.DataFrame()
            for key in data.keys():
                df[key] = pd.Series(data[key])

            keys = data.keys()
            if plot_val:
                keys = list(keys)[:int(len(keys)/2)]

            for key in keys:
                columns = [key]
                if plot_val:
                    columns.append(f"val_{key}")
                data=df[columns]
                data.index = range(1,len(data)+1) # Start at epoch 1
                plot = sns.lineplot(data=data, palette="tab10", linewidth=1.5)
                plot.set(yscale='log')
                plot.set(ylabel=key)
                plot.set(xlabel='epoch')
                plt.tight_layout()
                plt.savefig(f"figures/{name}/{key}")
                plt.figure()

def cmp_plot_history(names : "list[str]", plot_val=True):
    plots = {} # there will be as many plots as there are different metrics in the history files

    resulting_name = functools.reduce(lambda a, b : a + "_" + b, sorted(names)) # Compute output folder name

    # Ensure folder existence
    ensure_folders_exist([
        "figures",
        f"figures/cmp_{resulting_name}"
    ])

    # Parse histories and add to plot dict(key = metric)
    for name in names:
        with open(f'histories/{name}.json', 'r') as fp:
                data = json.load(fp)

                keys = data.keys()
                if plot_val:
                    keys = list(keys)[:int(len(keys)/2)]

                for key in keys:
                    # Ensure dataframe had been initialized
                    if key not in plots:
                        plots[key] = pd.DataFrame()

                    # Add data to dataframe
                    plots[key][name] = pd.Series(data[key])
                    if plot_val:
                        plots[key][f"val_{name}"] = pd.Series(data[f"val_{key}"])

    # Plot all metrics
    for key in plots.keys():
        plots[key].index = range(1,len(plots[key])+1) # Start at epoch 1
        plot = sns.lineplot(data=plots[key], palette="tab10", linewidth=1.5)
        plot.set(yscale='log')
        plot.set(ylabel=key)
        plot.set(xlabel='epoch')
        plt.tight_layout()
        plt.savefig(f"figures/cmp_{resulting_name}/{key}")
        plt.figure()