Created
August 3, 2019 07:39
-
-
Save tangh/0d398813dd3e64a72d830149c0363742 to your computer and use it in GitHub Desktop.
Show annotations in COCO dataset (multi-polygon and RLE format annos).
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from pycocotools.coco import COCO | |
import numpy as np | |
import cv2 | |
import os | |
coco_dataset_path = "/export/public/MS-COCO-2017/" | |
coco = COCO(coco_dataset_path + "annotations/instances_val2017.json") | |
# ============================================================================ | |
ann_ids = coco.getAnnIds(iscrowd=False) | |
anns = coco.loadAnns(ann_ids) | |
anns = [ann for ann in anns if len(ann['segmentation']) > 1] | |
print("num of annotations with more than one polygan:", len(anns)) # 3522 | |
for i, ann in enumerate(anns): | |
image_id = ann["image_id"] | |
segs = ann["segmentation"] | |
bbox = np.array(ann["bbox"]) | |
bbox[2:4] = bbox[0:2] + bbox[2:4] | |
print("BBox[{}] (xyxy):".format(i), bbox.tolist()) | |
image_info = coco.loadImgs(image_id) | |
image_path = image_info[0]["file_name"] | |
# [0] is required, always return a list | |
image_path = os.path.join(coco_dataset_path, "val2017", image_path) | |
print(image_path) | |
# /export/public/MS-COCO-2017/val2017/000000061108.jpg | |
image = cv2.imread(image_path) | |
segs = [np.array(seg, np.int32).reshape((1, -1, 2)) | |
for seg in segs] | |
for seg in segs: cv2.drawContours(image, seg, -1, (0,255,0), 2) | |
# third aug -1 means draw all contours in 3-D array, Or | |
# for seg in segs: cv2.fillPoly(image, segm, (0,255,0)) | |
cv2.rectangle(image, (int(bbox[0]), int(bbox[1])), | |
(int(bbox[2]), int(bbox[3])), (0,0,255), 2) | |
cv2.imshow("polygan label", image) | |
cv2.waitKey() | |
break | |
# ============================================================================ | |
ann_ids = coco.getAnnIds(iscrowd=True) | |
anns = coco.loadAnns(ann_ids) | |
import pycocotools.mask as mask | |
for ann in anns: | |
image_id = ann["image_id"] | |
segm = ann['segmentation'] | |
bbox = ann['bbox'] | |
assert sum(segm['counts']) == segm['size'][0] * segm['size'][1] | |
# Draw RLE label | |
label = np.zeros(segm['size'], np.uint8).reshape(-1) | |
ids = 0 | |
value = 0 | |
for c in segm['counts']: | |
label[ids: ids+c] = value | |
value = not value | |
ids += c | |
label = label.reshape(segm['size'], order='F') | |
# order='F' means Fortran memory order | |
cv2.imshow("RLE label", label*255) | |
cv2.waitKey() | |
print('Encoded RLE:', mask.frPyObjects(segm, *segm['size'])) | |
""" | |
Encoded RLE: {'size': [240, 320], 'counts': b'`824200N5OI?Y1o3]OfK@;T1m3]OkK\\O9W1k3^OnKXO7[1`3gNTLg0c0ZOYOX1n3IhL>V3DiL=U3FiL;V3HhL8W30cLO[34dLLZ36hLHX38hLHX39fLHZ38eLH\\38cLI]38bLG_3:`LF`3<ZLRN1a1f3U20]Oc0M3N2M3M3M3N2M3J6F:K5L4M3N2M3N2VOWJKk54ZJDj5:g0O1O100O10000000000000000O1000UJCc4=]KEa4;_KF`49aKI]48bKI]47cKK[45eKLZ44fKMY43fK0X40fK5W4LfKa0o3@nKc0Q4_OkKd0T4]OiKf0V4[OfKi0Y4XOcKk0]4VO^Ko0a4RO^Ko0a4QO^KP1b4QO]KP1b4PO]KQ1c4PO\\KP1d4PO\\KQ1c4oN\\KR1d4oN[KQ1e4oN[KQ1e4POYKQ1g4oNYKQ1g4j02J6O1kNWKCi4<ZK]OG]Oo4U1\\KZOHAl4U1^KjNF62Jj4V1gKnNb4m0dKQO]4n0S10kJoNk3Q1SLQOm3o0QLSOo3m0oKXOn3j0PKoNk0>o3h0PLWOQ4i0nKXOR4h0nKWOS4;WK_Oe06U4;XK]O5J14F1m4d0XKCLF`02]4d0XKL:_O_4e0WKL:^O`4f0VKM7]Oe4c0WK00@j4<\\K3H@n4=_KOBAR5?_KN^OAV5`0_KNh42ZKKg44]KFf49Y1O1O1O1O1O1O1000000O10000000000O10000001O00001O00001O001O2N1O2N1O2N4L8ZIVOV6R10000000001O001O0000]OQJFn5:VJBj5>XJ@h5`0YJ@f5`0[J_Oe5a0\\J_Oc5a0^J^Ob5b0_J^O`5b0`J_O_5a0bJ^O^5e0cJXO\\5j0d02N1O5dISOk5[1N1O001O001O1O001O00001O0000000000000000000000000000UOXJNh5MbJN^5OhJMY51kJMU51mJNT50nJ0R50oJNR52PKLP54V1002N1O2N1O2N2hI[OJ0X5f0fJXOJ560X5f0eJ@1LY5Y1bJkN]5e1O2N2N@hJiNW5g0ZKTO@HU5S1\\KTOj4k0WKTOj4k0WKUOi4k0WKTOj4m0UKSOk4m0UKUOi4k0XKVOf4j0P11O3M4L5oJaNg3_1XLcNg3^1WLdNi3[1VLfNk3[1PLiNP4Y1mKhNR4[1kKeNV4]1hKcNX4]1gKcNZ4]1eKcN\\4]1cKdN]4\\1aKeNa4[1]KeNd4`1UKaNm4P23M2N1O2N1O2N3M3M3M1O2M2O1N3N1N3N1N2O1N2N2N2N2N3L3N3L5J7HVi9'} | |
""" | |
image_info = coco.loadImgs(image_id) | |
image_path = image_info[0]["file_name"] | |
image_path = os.path.join(coco_dataset_path, "val2017", image_path) | |
print("image path (crowd label)", image_path) | |
# /export/public/MS-COCO-2017/val2017/000000448263.jpg | |
image = cv2.imread(image_path) | |
cv2.imshow("RLE label", image) | |
cv2.waitKey() | |
break |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Multi-Polygon Annos
RLE Annos