Added function to isolate images below 75kb

This commit is contained in:
2023-06-16 07:40:19 +03:00
parent b743bd30b6
commit 30dfe88aa1
+28 -1
View File
@@ -10,7 +10,31 @@ max_value = 255
threshold_type = cv2.THRESH_BINARY
def process_image(image_path,pass_folder,reject_folder):
def is_image_below_threshold(image_path, size_threshold):
"""If the image size is below the threshold, regardless of the thresholding
result, the image will be moved to the reject folder.
Args:
image_path (_type_): _description_
size_threshold (_type_): _description_
Returns:
_type_: _description_
"""
image_path = Path(image_path)
size = image_path.stat().st_size
return size < size_threshold
def process_image(image_path,pass_folder,reject_folder,size_threshold=75*1024):
"""_summary_
Args:
image_path (_type_): _description_
pass_folder (_type_): _description_
reject_folder (_type_): _description_
size_threshold (_type_, optional): _description_. Defaults to 75*1024.
"""
image_path = Path(image_path)
filename = image_path.name
image = cv2.imread(str(image_path))
@@ -25,4 +49,7 @@ def process_image(image_path,pass_folder,reject_folder):
else:
destination_folder = reject_folder
if is_image_below_threshold(image_path, size_threshold):
destination_folder = reject_folder
shutil.move(str(image_path), str(destination_folder / filename))