from keras.preprocessing import image
from keras.applications.vgg19 import preprocess_input
import numpy as np
import os
import uuid
from PIL import Image
def get_file_extension(filename):
if "." not in filename:
return ""
return filename.rsplit('.', 1)[1].lower()
def allowed_file(filename):
ALLOWED_EXTENSIONS = [
"png",
"jpg",
"jpeg",
]
return get_file_extension(filename) in ALLOWED_EXTENSIONS
def preprocess_image(image_path):
img = image.load_img(image_path, target_size=(224, 224))
img = image.img_to_array(img)
img = np.expand_dims(img, axis=0)
img = preprocess_input(img)
return img
def save_upload_file(original_name, file):
UPLOAD_DIR = "./dist/static/upload_images"
_, ext = os.path.splitext(original_name)
encrypted_name = str(uuid.uuid4()) + ext
print(encrypted_name)
file.save(os.path.join(UPLOAD_DIR, encrypted_name))
return encrypted_name
def get_image_scale(image_name):
path = f"../mm_images/{image_name}"
size = Image.open(path).size
return size[1] / size[0]
def re_arrange_images(image_names):
"""
因为css是纵向排列的，但一般要横着看，所以这里把图片重新排序。
(本来应该在前端写的，不过我不会写css，就代码改了)
"""
image_scales = [get_image_scale(name) for name in image_names]
columns = {}
scales = []
for i in range(5):
columns[i] = []
scales.append(0)
for i, name in enumerate(image_names):
which_col = scales.index(min(scales))
image_scale = image_scales[i]
columns[which_col].append(name)
scales[which_col] += image_scale
results = []
for i in range(5):
results += columns[i]
return results
