인프런 커뮤니티 질문&답변

안녕하세요 답변감사합니다 eval문제가 또발생한거같네요 WARNING:tensorflow:Detecting that an object or model or tf.train.Checkpoint is being deleted with unrestored values. See the following logs for the specific values in question. To silence these warnings, use status.expect_partial(). See https://www.tensorflow.org/api_docs/python/tf/train/Checkpoint#restorefor details about the status object returned by the restore function.

이런 문제가 떠서

import numpy as np

import skimage.io as io

from keras.preprocessing.image import ImageDataGenerator

from model import UNET_ISBI_2012

from absl import flags

from absl import app

import os

flags.DEFINE_string('checkpoint_path', default='saved_model_isbi_2012/unet_model.ckpt',

help='path to a directory to restore checkpoint file')

flags.DEFINE_string('test_dir', default='isbi_2012_test_result', help='directory which test prediction result saved')

flags.DEFINE_integer('num_classes', default=1, help='number of prediction classes')

FLAGS = flags.FLAGS

# set configuration value

batch_size = 1

total_test_image_num = 30

# normalize ISBI-2012 data

def normalize_isbi_2012(input_images):

# 0~255 -> 0.0~1.0

input_images = input_images / 255

return input_images

# make prediction mask

def create_mask(pred_mask):

pred_mask = np.where(pred_mask > 0.5, 1, 0)

return pred_mask[0]

# make test data generator

def make_test_generator(batch_size):

image_gen = ImageDataGenerator()

# set image and mask same augmentation using same seed

image_generator = image_gen.flow_from_directory(

directory='./isbi_2012/preprocessed',

classes=['test_imgs'],

class_mode=None,

target_size=(512, 512),

batch_size=batch_size,

color_mode='grayscale',

seed=1

)

for batch_images in image_generator:

batch_images = normalize_isbi_2012(batch_images)

yield batch_images

def main(_):

# check if checkpoint path exists

# if not os.path.exists(FLAGS.checkpoint_path):

# print('checkpoint file is not exists!')

# exit()

# create UNET model

unet_model = UNET_ISBI_2012(FLAGS.num_classes)

# restore latest checkpoint

status = unet_model.load_weights(FLAGS.checkpoint_path)

status.expect_partial() # Ignore warnings about incomplete restoration

print(f'{FLAGS.checkpoint_path} checkpoint is restored!')

# make generator

test_generator = make_test_generator(batch_size)

# check total image num

print('total test image :', total_test_image_num)

# save test prediction result to png file

if not os.path.exists(os.path.join(os.getcwd(), FLAGS.test_dir)):

os.mkdir(os.path.join(os.getcwd(), FLAGS.test_dir))

for image_num, test_image in enumerate(test_generator):

if image_num >= total_test_image_num:

break

pred_mask = unet_model.predict(test_image)

# Ensure the mask has three channels

output_image = create_mask(pred_mask)

output_image = np.stack((output_image,) * 3, axis=-1)

# Convert to uint8

output_image = (output_image * 255).astype(np.uint8)

output_image_path = os.path.join(os.getcwd(), FLAGS.test_dir, f'{image_num}_result.png')

io.imsave(output_image_path, output_image)

print(output_image_path + ' saved!')

if __name__ == '__main__':

app.run(main)

이렇게바꿧더니 결과사진이 이상하게 나오네요 ㅠ 맨위에 직선같은게 사진 결과입니다

(위사진 저장된 save file)

1. train부분 에러 train_isbi_2012.py:177: UserWarning: Model.fit_generator is deprecated and will be removed in a future version. Please use Model.fit, which supports generators. 이거를 Model.fit로 바꿔야하나요?

2. I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX AVX cpu로 돌아가고있는 뜻같은데 gpu를 따로쓰는방법이있나요?

1. evaluate 부분 에러 Found 30 images belonging to 1 classes.

1/1 [==============================] - 1s 955ms/step

evaluate_isbi_2012.py:85: UserWarning: C:\Users\Administrator\Desktop\UNET-tf2-main\isbi_2012_test_result\0_result.png is a low contrast image

io.imsave(output_image_path, create_mask(pred_mask))

Traceback (most recent call last):

File "evaluate_isbi_2012.py", line 89, in <module>

app.run(main)

File "D:\anaconda\envs\tfunet\lib\site-packages\absl\app.py", line 308, in run

runmain(main, args)

File "D:\anaconda\envs\tfunet\lib\site-packages\absl\app.py", line 254, in runmain

sys.exit(main(argv))

File "evaluate_isbi_2012.py", line 85, in main

io.imsave(output_image_path, create_mask(pred_mask))

File "D:\anaconda\envs\tfunet\lib\site-packages\skimage\io\_io.py", line 143, in imsave

return call_plugin('imsave', fname, arr, plugin=plugin, **plugin_args)

File "D:\anaconda\envs\tfunet\lib\site-packages\skimage\io\manage_plugins.py", line 207, in call_plugin

return func(*args, **kwargs)

File "D:\anaconda\envs\tfunet\lib\site-packages\imageio\v2.py", line 397, in imwrite

return file.write(im, **kwargs)

File "D:\anaconda\envs\tfunet\lib\site-packages\imageio\plugins\pillow.py", line 405, in write

raise ValueError("Can't write images with one color channel.")

ValueError: Can't write images with one color channel.

WARNING:tensorflow:Detecting that an object or model or tf.train.Checkpoint is being deleted with unrestored values. See the following logs for the specific values in question. To silence these warnings, use status.expect_partial(). See https://www.tensorflow.org/api_docs/python/tf/train/Checkpoint#restorefor details about the status object returned by the restore function.

W0510 18:14:50.947670 15764 checkpoint.py:205] Detecting that an object or model or tf.train.Checkpoint is being deleted with unrestored values. See the following logs for the specific values in question. To silence these warnings, use status.expect_partial(). See https://www.tensorflow.org/api_docs/python/tf/train/Checkpoint#restorefor details about the status object returned by the restore function.

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._iterations

W0510 18:14:50.952654 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._iterations

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._learning_rate

W0510 18:14:50.953642 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._learning_rate

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.1

W0510 18:14:50.954640 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._variables.1

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.2

W0510 18:14:50.955644 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._variables.2

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.3

W0510 18:14:50.955644 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._variables.3

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.4

W0510 18:14:50.956632 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._variables.4

WARNING:tensorflow:Value in checkpoint could not be found in the restored object: (root).optimizer._variables.5

W0510 18:14:50.957636 15764 checkpoint.py:214] Value in checkpoint could not be found in the restored object: (root).optimizer._variables.5

(optimizer._variables.96까지 에러,글자수 제한 있어 이렇게 씁니다)

안녕하세요 말하신데로 train evaluate 다바꿔서 5장 저장다하고 ckpt 로 교체해서 evaluate_isbi_2012.py 실행하니 이런문제가 뜨네요.감사합니다

안녕하세요~. 반갑습니다.

먼저 수강에 불편을 드려서 죄송합니다ㅠ.

해당 부분 관련해서 예전에 답변을 드렸던 내용이 있는데요.

https://www.inflearn.com/questions/687532/evaluate-isbi-2012-py-%EC%8B%A4%ED%96%89%EA%B4%80%EB%A0%A8-%EB%AC%B8%EC%9D%98%EB%93%9C%EB%A6%BD%EB%8B%88%EB%8B%A4

위에 내용을 정리하면 train_isbi_2012.py 파일과 evaluate_isbi_2012.py 파일들을 아래 파일들로 교체해서 실행해보시겠어요?

https://gist.github.com/solaris33/64feceb48079a2b339ef9f18a6544a74
https://gist.github.com/solaris33/771639041b8a4500b6d81951d4a2b814

첨부해주신 에러내용으로 추론해보면 현재 .pb 파일로 저장하도록 train_isbi_2012.py 파일을 수정했는데 evaluate_isbi_2012.py 파일에서는 .ckpt 파일을 불러오도록 수정해서 문제가 발생한것 같습니다.

위에 gist 파일들로 교체하시면 train_isbi_2012.py과 evaluate_isbi_2012.py 파일에서 모두 .ckpt 파일을 사용하게 됩니다.

해보시고 혹시 문제 있으면 다시 질문 주세요.

좋은하루 되세요.

감사합니다.