/
cocoextra.py
167 lines (127 loc) · 5.57 KB
/
cocoextra.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
import os
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image
import skimage.io as io
from pycocotools.coco import COCO
import json
# Extra utility COCO functions, not included in the default API
class COCOExtra:
def __init__(self, coco_ann, train_images_path):
self.coco_ann = coco_ann
self.train_images_path = train_images_path
# Overlay annotated category text directly on images
def ann_on_image(self, img_id, modified_ann_ids=None, save_fig=False, window_title=""):
coco_ann = self.coco_ann
img_info = coco_ann.loadImgs([img_id])[0]
img_file_name = img_info["file_name"]
image_path = os.path.join(self.train_images_path, img_file_name)
im = io.imread(image_path)
fig, ax = plt.subplots()
fig.canvas.set_window_title(window_title)
ax.axis("off")
ax.imshow(np.asarray(im))
ann_ids = coco_ann.getAnnIds(imgIds=[img_id], iscrowd=None)
anns = coco_ann.loadAnns(ann_ids)
coco_ann.showAnns(anns, draw_bbox=True)
cat_ids = coco_ann.getCatIds()
for i, ann in enumerate(anns):
entity_id = anns[i]["category_id"]
if entity_id not in cat_ids:
continue
entity = coco_ann.loadCats(entity_id)[0]["name"]
if modified_ann_ids and anns[i]["id"] in modified_ann_ids:
facecolor = 'red'
else:
facecolor = 'white'
ax.text(anns[i]['bbox'][0], anns[i]['bbox'][1], entity, style='italic',
bbox={'facecolor': facecolor, 'alpha': 0.7, 'pad': 5})
if save_fig:
if modified_ann_ids:
plt.savefig(f"{img_id}_annotated_injected.jpg", bbox_inches="tight", pad_inches=0)
else:
plt.savefig(f"{img_id}_annotated.jpg", bbox_inches="tight", pad_inches=0)
else:
plt.show()
# Overlay faulty annotations directly on the image (right),
# side by side with the original annotated version of the image (left)
def ann_on_image_sidebyside(self, img_id, coco_ann_orig, modified_ann_ids=None, save_fig=False, window_title=""):
coco_ann = self.coco_ann
img_info = coco_ann.loadImgs([img_id])[0]
img_file_name = img_info["file_name"]
image_path = os.path.join(self.train_images_path, img_file_name)
im_orig = io.imread(image_path)
im = io.imread(image_path)
fig, ax = plt.subplots(1, 2)
fig.canvas.set_window_title(window_title)
ax[0].axis("off")
ax[1].axis("off")
ax[0].set_title("Before")
ax[1].set_title("After")
ax[0].imshow(np.asarray(im_orig))
ax[1].imshow(np.asarray(im))
# Orig annotations
orig_ann_ids = coco_ann_orig.getAnnIds(imgIds=[img_id], iscrowd=None)
orig_anns = coco_ann_orig.loadAnns(orig_ann_ids)
coco_ann_orig.showAnns(orig_anns, draw_bbox=True, ax_select=ax[0])
cat_ids = coco_ann_orig.getCatIds()
for i, ann in enumerate(orig_anns):
entity_id = orig_anns[i]["category_id"]
if entity_id not in cat_ids:
continue
entity = coco_ann.loadCats(entity_id)[0]["name"]
ax[0].text(orig_anns[i]['bbox'][0], orig_anns[i]['bbox'][1], entity, style='italic',
bbox={'facecolor': 'white', 'alpha': 0.7, 'pad': 5})
# Faulty annotations
ann_ids = coco_ann.getAnnIds(imgIds=[img_id], iscrowd=None)
anns = coco_ann.loadAnns(ann_ids)
coco_ann.showAnns(anns, draw_bbox=True)
cat_ids = coco_ann.getCatIds()
for i, ann in enumerate(anns):
entity_id = anns[i]["category_id"]
if entity_id not in cat_ids:
continue
entity = coco_ann.loadCats(entity_id)[0]["name"]
if modified_ann_ids and anns[i]["id"] in modified_ann_ids:
facecolor = 'red'
else:
facecolor = 'white'
ax[1].text(anns[i]['bbox'][0], anns[i]['bbox'][1], entity, style='italic',
bbox={'facecolor': facecolor, 'alpha': 0.7, 'pad': 5})
if save_fig:
if modified_ann_ids:
plt.savefig(f"{img_id}_annotated_injected_sidebyside.jpg", bbox_inches="tight", pad_inches=0)
else:
plt.savefig(f"{img_id}_annotated.jpg", bbox_inches="tight", pad_inches=0)
else:
plt.show()
# Display image by image id
def show_img(self, img_id, save_fig=False):
coco_ann = self.coco_ann
img_info = coco_ann.loadImgs([img_id])[0]
img_file_name = img_info["file_name"]
image_path = os.path.join(self.train_images_path, img_file_name)
im = io.imread(image_path)
plt.axis("off")
plt.imshow(np.asarray(im))
if save_fig:
plt.savefig(f"{img_id}.jpg", bbox_inches="tight", pad_inches=0)
else:
plt.show()
# Save modified annotation file
def save_ann(self, filename):
with open(filename, "w") as outfile:
json.dump(self.coco_ann.dataset, outfile, indent=4)
# Get dict: {supercategory: [category_ids]}
def get_supercat_dict(self):
coco_ann = self.coco_ann
allcats = coco_ann.cats
result = {}
for catId, cat in allcats.items():
supercat = cat["supercategory"]
if not supercat:
raise ValueError("Dataset has no supercategories!")
if supercat not in result:
result[supercat] = []
result[supercat].append(catId)
return result