def create_image_hash(image_dir):
return {
image_name:
f"{get_image_average_hash(image_dir / image_name, 10)}{os.path.splitext(image_name)[-1]}"
for image_name in os.listdir(image_dir)
if check_if_approved_image_extension(image_name)
}
def migrate_images(dataset_dir, raw_dir, image_hash):
image_dir = dataset_dir / "images"
mkdir_safely(image_dir)
for image_origin in os.listdir(raw_dir):
if check_if_approved_image_extension(image_origin):
image_path = dataset_dir / "images" / image_hash[image_origin]
if not os.path.isfile(image_path):
copyfile(raw_dir / image_origin, image_path)
def migrate_images(dataset_dir, raw_dir, image_hash):
image_dir = dataset_dir / "images"
mkdir_safely(image_dir)
for image_name_origin in os.listdir(raw_dir):
if check_if_approved_image_extension(image_name_origin):
image_path = dataset_dir / "images" / image_hash[image_name_origin]
if not os.path.isfile(image_path):
image_origin = Image.open(raw_dir / image_name_origin)
image_without_exif = Image.new(image_origin.mode, image_origin.size)
image_without_exif.putdata(image_origin.getdata())
image_without_exif.save(image_path)
def thumbnailing(thumbnails_dir, repository_dir, THUMBNAIL_SIZE=(300, 300)):
for dirpath, dirnames, filenames in os.walk(repository_dir):
for filename in filenames:
if check_if_approved_image_extension(filename):
image_path = thumbnails_dir / filename[:2] / filename
if not image_path.is_file():
image = Image.open(f"{dirpath}/{filename}")
if check_if_approved_image_format(image.format):
image.thumbnail(THUMBNAIL_SIZE)
if filename[:2] not in os.listdir(thumbnails_dir):
os.mkdir(thumbnails_dir / filename[:2])
image.save(image_path)
def thumbnailing(thumbnails_dir,
repository_dir,
weedcoco_path,
THUMBNAIL_SIZE=(300, 300)):
with open(weedcoco_path) as f:
coco = json.load(f)
denormalise_weedcoco(coco)
coco_by_filename = {
os.path.basename(image["file_name"]): image
for image in coco["images"]
}
for dirpath, dirnames, filenames in os.walk(repository_dir):
for filename in filenames:
if not check_if_approved_image_extension(filename):
# XXX: should this raise an error?
continue
thumbnail_one(
coco_by_filename[filename],
f"{dirpath}/{filename}",
thumbnails_dir,
thumbnail_size=THUMBNAIL_SIZE,
)
import argparse
import pathlib
import os
import scipy.spatial
import PIL.Image
import seaborn as sns
import matplotlib.pyplot as plt
from weedcoco.utils import check_if_approved_image_extension
sns.set_theme(color_codes=True)
ap = argparse.ArgumentParser(description=__doc__)
ap.add_argument("--dataset-dir", required=True, type=pathlib.Path)
args = ap.parse_args()
X = np.asarray(
[
imagehash.average_hash(
PIL.Image.open(f"{dirpath}/{filename}"), hash_size=16
).hash.flatten()
for dirpath, dirnames, filenames in os.walk(args.dataset_dir)
if os.path.basename(dirpath) == "img"
for filename in filenames
if check_if_approved_image_extension(filename)
]
)
distances = scipy.spatial.distance.pdist(X, metric="hamming")
g = sns.displot(distances)
plt.show()