ValueError: Error when checking input: expected input_1 to have 4 dimensions,

Question

前提・実現したいこと

以下のコードを実行したところエラーが出てしまいました。ご教授お願いします。

発生している問題・エラーメッセージ

Traceback (most recent call last):
  File "sleepingmodel6.py", line 106, in <module>
    history = model.fit_generator(train_gen,steps_per_epoch=len(train_idx)//batch_size,epochs=10,callbacks=callbacks,validation_data=valid_gen,validation_steps=len(valid_idx)//valid_batch_size)
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/engine/training.py", line 1732, in fit_generator
    initial_epoch=initial_epoch)
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/engine/training_generator.py", line 220, in fit_generator
    reset_metrics=False)
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/engine/training.py", line 1508, in train_on_batch
    class_weight=class_weight)
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/engine/training.py", line 579, in _standardize_user_data
    exception_prefix='input')
  File "/opt/anaconda3/lib/python3.7/site-packages/keras/engine/training_utils.py", line 135, in standardize_input_data
    'with shape ' + str(data_shape))
ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (32, 198, 198)

該当のソースコード

import numpy as np
import pandas as pd

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
    for filename in filenames:
        print(os.path.join(dirname, filename))

import glob
import matplotlib.pyplot as plt
import seaborn as sns
from PIL import Image
from efficientnet.keras import EfficientNetB7  
from sklearn.model_selection import train_test_split
from keras.utils import to_categorical
from keras.layers import Input, Dense
from keras.models import Model
from keras.callbacks import ModelCheckpoint

DATA_DIR = 's0001/'
IM_WIDTH = IM_HEIGHT = 198
TRAIN_TEST_SPLIT = 0.8
TRAIN_VALID_SPLIT = 0.7
ID_EYE_MAP = {0: 'close', 1: 'open'}
EYE_ID_MAP = dict((r, i) for i, r in ID_EYE_MAP.items())

def parse_filepath(filepath):
    try:
        path, filename = os.path.split(filepath)
        filename, ext = os.path.splitext(filename)
        _, _, _, _, eyestatus, _, _, _, = filename.split("_")
        return ID_EYE_MAP[int(eyestatus)]

    except Exception as e:
        print(filepath)
        return None, None, None

files = glob.glob(os.path.join(DATA_DIR, "*.png"))
attributes = list(map(parse_filepath, files))

df = pd.DataFrame(attributes)
df['file'] = files
df.columns = ['eyestatus', 'file']
df = df.dropna()
df['eyestatus_id'] = df['eyestatus'].map(lambda eyestatus: EYE_ID_MAP[eyestatus])

#print(df.isnull().sum())
print(df)
p = np.random.permutation(len(df))
train_up_to = int(len(df) * TRAIN_TEST_SPLIT)
train_idx = p[:train_up_to]
test_idx = p[train_up_to:]
train_up_to = int(train_up_to * TRAIN_VALID_SPLIT)
train_idx, valid_idx = train_idx[:train_up_to], train_idx[train_up_to:]

def get_data_generator(df, indices, for_training, batch_size=32):

    images, eyestatuses = [], []
    while True:
        for i in indices:
            r = df.iloc[i]
            file, eyestatus = r['file'], r['eyestatus_id']
            im = Image.open(file)
            im = im.resize((IM_WIDTH, IM_HEIGHT))
            im = np.array(im) / 255.0
            images.append(im)
            eyestatuses.append(to_categorical(eyestatus, 2))
            if len(images) >= batch_size:
                yield np.array(images), [np.array(eyestatuses)]
                images, eyestatuses = [], []
        if not for_training:
            break

input_layer = Input(shape=(IM_HEIGHT, IM_WIDTH, 3))
#input_layer = Reshape((IM_HEIGHT, IM_WIDTH, 3), input_shape = (IM_HEIGHT, IM_WIDTH))(F)

efficient_net = EfficientNetB7(weights='noisy-student',include_top=False,input_tensor = input_layer,pooling='max')

for layer in efficient_net.layers:
    layer.trainable = True

bottleneck=efficient_net.output

x = Dense(units=128, activation='relu')(bottleneck)
eyestatus_output = Dense(units=len(EYE_ID_MAP), activation='softmax', name='eyestatus_output')(x)

model = Model(inputs=input_layer, outputs=[eyestatus_output])

model.compile(optimizer='rmsprop',loss={'eyestatus_output': 'categorical_crossentropy'},loss_weights={'eyestatus_output': 1.},metrics={'eyestatus_output': 'accuracy'})

#model.summary()

batch_size = 32
valid_batch_size = 32

train_gen = get_data_generator(df, train_idx, for_training=True, batch_size=batch_size)
valid_gen = get_data_generator(df, valid_idx, for_training=True, batch_size=valid_batch_size)

callbacks = [
    ModelCheckpoint('./model_checkpoint', monitor='val_loss', verbose=1, save_best_only=True, mode='min')
]

history = model.fit_generator(train_gen,steps_per_epoch=len(train_idx)//batch_size,epochs=10,callbacks=callbacks,validation_data=valid_gen,validation_steps=len(valid_idx)//valid_batch_size)

試したこと

次元数の部分でエラーが起きていることはわかるのですが、どう対処すれば良いかわかりません。

補足情報（FW/ツールのバージョンなど）

ここにより詳細な情報を記載してください。

Answer 1

本来(32, 198, 198, 3)の3ch画像の入力が想定されるところに、(32, 198, 198)の1ch画像が入力されているのが原因だと考えられます。コード自体は、特に問題ないように思うのですが、入力画像はグレースケール画像だったりしないでしょうか？
グレースケール画像の場合は、get_data_generatorを

def get_data_generator(df, indices, for_training, batch_size=32):

    images, eyestatuses = [], []
    while True:
        for i in indices:
            r = df.iloc[i]
            file, eyestatus = r['file'], r['eyestatus_id']
            im = Image.open(file)
            im = im.convert("RGB") #<-RGB化
            im = im.resize((IM_WIDTH, IM_HEIGHT))
            im = np.array(im) / 255.0
            images.append(im)
            eyestatuses.append(to_categorical(eyestatus, 2))
            if len(images) >= batch_size:
                yield np.array(images), [np.array(eyestatuses)]
                images, eyestatuses = [], []
        if not for_training:
            break

と修正することで通るはずです。
ざっと見る限り、それ以外にざっと問題点はなさそうですが…、見落としていたら申し訳ありません。inputのchannelがらみのエラーなのは間違いないと思うので、もし解決しない場合はget_data_generatorのコード内にprint(im.shape)を挿入してshapeを確認していくと、どこで問題が発生しているのかわかると思います。

Comments

[c7744kumamoto-u]

入力がグレースケール画像でした、、、
指摘通りにしたところ、無事解決いたしました。ありがとうございます！