-
Notifications
You must be signed in to change notification settings - Fork 3
/
evaluate.py
204 lines (178 loc) · 6.92 KB
/
evaluate.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
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
198
199
200
201
202
203
204
from argparse import ArgumentParser
import re
import os
from tqdm import tqdm
import numpy as np
import cv2
import json
from ji import init, process_image
def valid_gt_text(text):
return not (re.search(r"[\WA-Za-z]", text)) and len(text) >= 2
def get_text_and_box(labels, verify=False):
texts = list()
boxes = list()
for label in labels:
text = label[-1]
if verify and not valid_gt_text(text):
continue
texts.append(text)
box = label[:8]
box = [int(point) for point in box]
boxes.append(box)
return texts, boxes
def get_label_lines(path, verify=False):
box_text = list()
with open(path, encoding='utf-8') as file:
lines = file.readlines()
for line in lines:
text = line.rstrip().split(",")[-1]
if verify and not valid_gt_text(text):
continue
box = line.rstrip().split(",")[:8]
box = [int(point) for point in box]
box.append(text)
box_text.append(box)
return box_text
def match_text(text1, text2):
match = 0
for i in range(len(text2)):
index = text1.index(text2[i]) if text2[i] in text1 else -1
if index != -1:
match += 1
text1 = text1[index + 1:]
return match
def count_similarity(text1, text2):
if text1 == text2:
return 1
match1 = match_text(text1, text2)
match2 = match_text(text2, text1)
match = max(match1, match2)
return match / (len(text1) + len(text2) - match)
def cal_iou(box1, box2):
"""
box1, box2: list or numpy array of size 4*2 or 8, h_index first
"""
box1 = np.array(box1).reshape([1, 4, 2])
box2 = np.array(box2).reshape([1, 4, 2])
box1_max = box1.max(axis=1)
box2_max = box2.max(axis=1)
w_max = max(box1_max[0][0], box2_max[0][0])
h_max = max(box1_max[0][1], box2_max[0][1])
canvas = np.zeros((h_max + 1, w_max + 1))
box1_canvas = canvas.copy()
box1_area = np.sum(cv2.drawContours(
box1_canvas, box1, -1, 1, thickness=-1))
box2_canvas = canvas.copy()
box2_area = np.sum(cv2.drawContours(
box2_canvas, box2, -1, 1, thickness=-1))
cv2.drawContours(canvas, box1, -1, 1, thickness=-1)
cv2.drawContours(canvas, box2, -1, 1, thickness=-1)
union = np.sum(canvas)
intersction = box1_area + box2_area - union
return intersction / union
def evaluate(pred_json, gt_json, threshold_text, threshold_box):
pred_dict = json.loads(pred_json)
gt_dict = json.loads(gt_json)
assert gt_dict.keys() == pred_dict.keys()
img_ids = gt_dict.keys()
correct_num = 0
pred_num = 0
gt_num = 0
pbar = tqdm(enumerate(img_ids), total=len(img_ids), desc="Eval")
for index, img_id in pbar:
pbar.set_description("Evaluation - i:{}, {}".format(index + 1, img_id))
gt_texts, gt_boxes = get_text_and_box(gt_dict[img_id], verify=True)
pred_texts, pred_boxes = get_text_and_box(pred_dict[img_id])
gt_num += len(gt_texts)
pred_num += len(pred_texts)
matched_i = list()
matched_j = list()
for i in range(len(gt_texts)):
for j in range(len(pred_texts)):
if i in matched_i or j in matched_j:
continue
score_text = count_similarity(gt_texts[i], pred_texts[j])
score_bbox = cal_iou(gt_boxes[i], pred_boxes[j])
if score_text >= threshold_text and score_bbox >= threshold_box: # correct prediction
correct_num += 1
matched_i.append(i)
matched_j.append(j)
pbar.close()
# overall result
acc = round(correct_num / pred_num, 4) if pred_num else 0
recall = round(correct_num / gt_num, 4) if gt_num else 0
lines = list()
lines.append("\n-------------------Evaluation-------------------")
lines.append("gt_num: {}".format(gt_num))
lines.append("pred_num: {}".format(pred_num))
lines.append("correct_num: {}".format(correct_num))
lines.append("acc: {}".format(acc))
lines.append("recall: {}\n".format(recall))
lines = "\n".join(lines)
print(lines)
return acc, recall
def get_gt_json(gt_dir):
filenames = os.listdir(gt_dir)
filenames = [filename for filename in filenames if ".txt" in filename]
gt_paths = [os.path.join(gt_dir, filename) for filename in filenames]
img_ids = [filename.split(".")[0] for filename in filenames]
gt_dict = dict()
pbar = tqdm(enumerate(img_ids), total=len(img_ids), desc="GET LABELS")
for index, img_id in pbar:
pbar.set_description("GET LABELS - i:{}, {}".format(index + 1, img_id))
gt_path = gt_paths[index]
gt_labels = get_label_lines(gt_path)
gt_dict[img_id] = gt_labels
gt_json = json.dumps(gt_dict)
return gt_json
def get_pred_json(img_dir):
model = init()
filenames = os.listdir(img_dir)
filenames = [filename for filename in filenames if ".jpg" in filename]
img_paths = [os.path.join(img_dir, filename) for filename in filenames]
img_ids = [filename.split(".")[0] for filename in filenames]
pred_dict = dict()
pbar = tqdm(enumerate(img_ids), total=len(img_ids), desc="Prediction")
for index, img_id in pbar:
pbar.set_description("Prediction - i:{}, {}".format(index + 1, img_id))
img_path = img_paths[index]
input_image = cv2.imread(img_path)
img_pred_json = process_image(model, input_image, args=None)
img_pred = json.loads(img_pred_json)
pred_dict[img_id] = [i["points"] + [i["name"]]
for i in img_pred["model_data"]["objects"]]
pred_json = json.dumps(pred_dict)
return pred_json
def main():
gt_json = get_gt_json(gt_dir=args.gt)
pred_json = get_pred_json(img_dir=args.img)
acc, recall = evaluate(
pred_json, gt_json, threshold_text=args.threshold_text, threshold_box=args.threshold_box)
if __name__ == "__main__":
parser = ArgumentParser()
parser.add_argument("-i",
"--img",
default="/home/data/1411",
type=str,
help="path of images.",
metavar="IMAGE PATH")
parser.add_argument("-g",
"--gt",
default="/home/data/1411",
type=str,
help="path of ground truth.",
metavar="GT PATH")
parser.add_argument("-t1",
"--threshold_text",
default=0.6,
type=float,
help="threshold for evaluate text.",
metavar="THRESHOLD TEXT")
parser.add_argument("-t2",
"--threshold_box",
default=0.5,
type=float,
help="threshold for evaluate bounding box.",
metavar="THRESHOLD BOUNDING BOX")
args = parser.parse_args()
main()