Datasets:
File size: 1,383 Bytes
ada18d0 | 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 | import os
import random
import sys
import matplotlib.pyplot as plt
from PIL import Image
from pycocotools.coco import COCO
coco_folder = "instance_segmentation/multiannotator_15imgs_coco"
img_folder = "images"
if __name__ == "__main__":
# Optional user id as argument, otherwise pick a random user
user_ids = sorted(f[:-len(".json")] for f in os.listdir(coco_folder) if f.endswith(".json"))
user_id = sys.argv[1] if len(sys.argv) > 1 else random.choice(user_ids)
coco = COCO(os.path.join(coco_folder, f"{user_id}.json"))
img_ids = sorted(coco.imgs)
print(f"User {user_id}: {len(img_ids)} images, {len(coco.anns)} annotations")
cols = min(3, len(img_ids))
rows = -(-len(img_ids) // cols)
fig, axes = plt.subplots(rows, cols, figsize=(6 * cols, 3.6 * rows), squeeze=False)
for ax in axes.flat:
ax.axis("off")
for ax, img_id in zip(axes.flat, img_ids):
img_info = coco.imgs[img_id]
anns = coco.loadAnns(coco.getAnnIds(imgIds=img_id))
ax.imshow(Image.open(os.path.join(img_folder, img_info["file_name"])))
plt.sca(ax)
coco.showAnns(anns, draw_bbox=False)
ax.set_title(f"{img_info['file_name']} ({len(anns)} annotations)", fontsize=9)
fig.suptitle(f"Annotations of user {user_id}")
fig.tight_layout()
fig.savefig("example_user_annotations.png", dpi=100)
plt.show()
|