mirror of
https://github.com/azaion/ai-training.git
synced 2026-04-23 03:06:35 +00:00
+36
-54
@@ -1,15 +1,20 @@
|
||||
import os.path
|
||||
import albumentations as A
|
||||
import time
|
||||
|
||||
import cv2
|
||||
import albumentations as alb
|
||||
from os import listdir
|
||||
from os.path import isfile, join
|
||||
from pathlib import Path
|
||||
|
||||
labels_dir = 'labels'
|
||||
images_dir = 'images'
|
||||
current_dataset_dir = os.path.join('datasets', 'zombobase-current')
|
||||
|
||||
class ImageAnnotation:
|
||||
|
||||
def read_annotations(self) -> [[]]:
|
||||
with open(self.dataset_annotation_path, 'r') as f:
|
||||
with open(self.annotation_path, 'r') as f:
|
||||
rows = f.readlines()
|
||||
arr = []
|
||||
for row in rows:
|
||||
@@ -18,73 +23,50 @@ class ImageAnnotation:
|
||||
coordinates = [float(n) for n in str_coordinates]
|
||||
coordinates.append(class_num)
|
||||
arr.append(coordinates)
|
||||
|
||||
return arr
|
||||
|
||||
def __init__(self, image_path):
|
||||
self.image_path = image_path
|
||||
self.image_name = Path(image_path).stem
|
||||
self.dataset_image_path = os.path.join(current_dataset_dir, images_dir, self.image_path + '.jpg')
|
||||
self.image = cv2.imread(self.dataset_image_path)
|
||||
self.dataset_image_path = os.path.join(current_dataset_dir, images_dir, self.image_name, os.path.basename(image_path))
|
||||
self.image = cv2.imread(image_path)
|
||||
|
||||
self.annotation_path = os.path.join(labels_dir, self.image_path + '.txt')
|
||||
self.dataset_annotation_path = os.path.join(current_dataset_dir, labels_dir, self.image_path + '.txt')
|
||||
self.annotation_path = os.path.join(labels_dir, self.image_name, '.txt')
|
||||
self.dataset_annotation_path = os.path.join(current_dataset_dir, labels_dir, self.image_name, '.txt')
|
||||
self.annotations = self.read_annotations()
|
||||
|
||||
|
||||
def image_processing(img_ann: ImageAnnotation) -> [ImageAnnotation]:
|
||||
category_ids = []
|
||||
bboxes = ImageAnnotation(img_ann).read_annotations()
|
||||
for i in range(len(bboxes)):
|
||||
category_ids.append(bboxes[i][4])
|
||||
bboxes[i].pop(4)
|
||||
|
||||
transform = A.Compose([
|
||||
A.HorizontalFlip(p=0.5),
|
||||
A.RandomBrightnessContrast(p=0.2),
|
||||
A.ShiftScaleRotate(shift_limit=0.1, scale_limit=0.2, rotate_limit=15, p=0.5),
|
||||
], bbox_params=A.BboxParams(format='yolo', label_fields=['category_ids']))
|
||||
|
||||
bboxes = bboxes
|
||||
imag = ImageAnnotation(img_ann).image
|
||||
|
||||
transformed = transform(image=imag, bboxes=bboxes, category_ids=category_ids)
|
||||
transformed_image = transformed['image']
|
||||
transformed_bboxes = transformed['bboxes']
|
||||
transformed_category_ids = transformed['category_ids']
|
||||
|
||||
return transformed_image, transformed_bboxes, transformed_category_ids
|
||||
# return structure example:
|
||||
# utilize transform albumentations here
|
||||
return [ImageAnnotation(f'{img_ann.image_name}1', image1, bboxes1 ),
|
||||
ImageAnnotation(f'{img_ann.image_name}2', image2, bboxes2),
|
||||
...
|
||||
]
|
||||
|
||||
def write_results(img_ann: ImageAnnotation):
|
||||
for i in range(10):
|
||||
transformed_image, transformed_bboxes, transformed_category_ids, = image_processing(img_ann)
|
||||
cv2.imwrite(os.path.join(current_dataset_dir, images_dir, str(i)+ImageAnnotation(img_ann).image_path + '.jpg'), transformed_image)
|
||||
with open(os.path.join(current_dataset_dir, labels_dir, str(i)+ImageAnnotation(img_ann).image_path + '.txt'), 'w') as f:
|
||||
for bbox, category_id in zip(transformed_bboxes, transformed_category_ids):
|
||||
x_center, y_center, width, height = bbox
|
||||
cla = category_id
|
||||
f.write(f"{cla} {x_center} {y_center} {width} {height}\n")
|
||||
#
|
||||
#
|
||||
def process_image():
|
||||
file_annotation = []
|
||||
file_annotation_finished =[]
|
||||
for foldername, subfolders, filenames in os.walk(os.path.join(current_dataset_dir,images_dir)):
|
||||
file_annotation.append(filenames)
|
||||
print(file_annotation)
|
||||
# write image cv2.imwrite(, image) dataset_image_path
|
||||
# write img_ann.annotations into new file with name dataset_annotation_path
|
||||
|
||||
for i in range(len(file_annotation[0])):
|
||||
ff = file_annotation[0][i].split('.')
|
||||
ff.pop(-1)
|
||||
ff = '.'.join(ff)
|
||||
file_annotation_finished.append(ff)
|
||||
|
||||
return file_annotation_finished
|
||||
def process_image(img_ann):
|
||||
results = image_processing(img_ann)
|
||||
for res_ann in results:
|
||||
write_results(res_ann)
|
||||
write_results(img_ann)
|
||||
os.remove(img_ann.image_path)
|
||||
os.remove(img_ann.annotation_path)
|
||||
|
||||
|
||||
def main():
|
||||
process_image()
|
||||
for i in process_image():
|
||||
write_results(i)
|
||||
while True:
|
||||
images = os.listdir(images_dir)
|
||||
if len(images) == 0:
|
||||
time.sleep(5)
|
||||
continue
|
||||
|
||||
for image in images:
|
||||
image_path = os.path.join(images_dir, image)
|
||||
process_image(ImageAnnotation(image_path))
|
||||
|
||||
if __name__ == '__main__':
|
||||
main()
|
||||
Reference in New Issue
Block a user