-
Notifications
You must be signed in to change notification settings - Fork 0
/
iou.py
139 lines (94 loc) · 3.93 KB
/
iou.py
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
# -*- coding: utf-8 -*-
"""
Created on Fri Apr 5 11:35:27 2019
@author: Jrainbow
"""
from skimage.io import imread
import numpy as np
import pandas as pd
groundtruth_raster_path_road = r"U:\EOResearch\Ecopia\Topo\AllRasterized\topo_rasterized_road_clipped.tif"
comparison_raster_path_road = r"U:\EOResearch\Ecopia\Aerial_27700\AllRasterized\ecopia_rasterized_road_clipped.tif"
groundtruth_raster_path_building = r"U:\EOResearch\Ecopia\Topo\AllRasterized\topo_rasterized_building_clipped.tif"
comparison_raster_path_building = r"U:\EOResearch\Ecopia\Aerial_27700\AllRasterized\ecopia_rasterized_building_clipped.tif"
#image_dict = {"buildings": [groundtruth_raster_path_building, comparison_raster_path_building],
# "roads": [groundtruth_raster_path_road, comparison_raster_path_road]}
image_dict = {"buildings": [comparison_raster_path_building, groundtruth_raster_path_building],
"roads": [comparison_raster_path_road, groundtruth_raster_path_road]}
def normalize(img):
"""
Converts an image into binary form (1 when class is present)
Input: img array
Output: binary img array
"""
mini = img.min()
# Set negatives to 0
img[img == mini] = 0
# Set positives to 1
img[img > mini] = 1
out = img.astype(np.uint8)
return out
def calc_iou(groundtruth, comparison):
"""
Calculates the Intersection over Union (Jacard Index) of two raster images
Inputs: groundtruth raster image, comparison raster image
Outputs: Binarized ground truth image, binarized comparison image, IoU score
"""
gt = imread(groundtruth, as_gray=True)
cp = imread(comparison, as_gray=True)
gt_norm = normalize(gt)
cp_norm = normalize(cp)
a = gt_norm & cp_norm
b = gt_norm | cp_norm
anz = np.count_nonzero(a)
bnz = np.count_nonzero(b)
iou = anz / bnz
print("Intersection over Union is {}%".format(iou * 100))
return gt_norm, cp_norm, iou
def raw_matrix(image_dictionary):
"""
Calculates the pixel-wise True Positives, True Negatives, False Positives, False Negatives of two raster images
Input: Dictionary of classes, e.g.
image_dict = {"buildings": [comparison_raster_path_building, groundtruth_raster_path_building],
"roads": [comparison_raster_path_road, groundtruth_raster_path_road]}
Output: Dataframe
"""
tp = []
fp = []
tn = []
fn = []
for clz in image_dictionary:
gt = imread(image_dictionary[clz][0], as_gray=True)
cp = imread(image_dictionary[clz][1], as_gray=True)
gt_norm = normalize(gt)
cp_norm = normalize(cp)
gt_inv = 1 - gt_norm
cp_inv = 1 - cp_norm
a = gt_norm & cp_norm
b = gt_inv & cp_norm
c = cp_inv & gt_norm
d = cp_inv & gt_inv
anz = np.count_nonzero(a)
bnz = np.count_nonzero(b)
cnz = np.count_nonzero(c)
dnz = np.count_nonzero(d)
tp.append(anz)
fp.append(bnz)
fn.append(cnz)
tn.append(dnz)
results = pd.DataFrame({"Class:": list(image_dictionary.keys()),
"True Positives": tp,
"False Positives": fp,
"True Negatives": tn,
"False Negatives": fn})
return results
def accuracy_metrics(raw):
"""
Adds Recall, Precision and F1 scroes to the raw_matrix above
Input: raw_matrix
Output: same, with extra columns for Recall, Precision and F1
"""
raw['Recall'] = raw['True Positives'] / (raw['True Positives'] + raw['False Negatives'])
raw['Precision'] = raw['True Positives'] / (raw['True Positives'] + raw['False Positives'])
raw['F1'] = 2 * (raw['Precision'] * raw['Recall']) / (raw['Precision'] + raw['Recall'])
return raw
raw = raw_matrix(image_dict)