Tensorflow.Keras 说输入与预期输入不同

Question

I am writing a simple CNN to classify different features of a cartoon face. I am using this dataset. When I try running my code, I get the following error:

Traceback (most recent call last):
  File "cartoonKerasPy.py", line 86, in <module>
    model.fit(x_train, y_train, batch_size=8, epochs=3)
  File "C:\Users\befbr\Anaconda3\envs\Cartoon AI\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1536, in fit
    validation_split=validation_split)
  File "C:\Users\befbr\Anaconda3\envs\Cartoon AI\lib\site-packages\tensorflow\python\keras\engine\training.py", line 992, in _standardize_user_data
    class_weight, batch_size)
  File "C:\Users\befbr\Anaconda3\envs\Cartoon AI\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1154, in _standardize_weights
    exception_prefix='target')
  File "C:\Users\befbr\Anaconda3\envs\Cartoon AI\lib\site-packages\tensorflow\python\keras\engine\training_utils.py", line 293, in standardize_input_data
    str(len(data)) + ' arrays: ' + str(data)[:200] + '...')
ValueError: Error when checking model target: the list of Numpy arrays that you are passing to your model is not the size the model expected. Expected to see 1 array(s), but instead got the following list of 160 arrays: [0, 0, 0, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0...

Here is my code:

print("Importing Packages")
import os
import tensorflow as tf
from tensorflow import keras as k
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import numpy as np
import pandas as pd
from PIL import Image

def createModel():
    model = k.models.Sequential([
      k.layers.Conv3D(filters=64, kernel_size=2, padding='same', activation='relu', input_shape=(500,500, 4, 1)),
      k.layers.MaxPooling3D(pool_size=2),
      k.layers.Dropout(0.2),
      k.layers.Flatten(),
      k.layers.Dense(1000, activation='relu'),
      k.layers.Dense(1, activation='softmax')

    ])
    model.compile(optimizer='adam',
                  loss='sparse_categorical_crossentropy',
                  metrics=['accuracy'])
    return model

def getFile(dir, getEveryNthLine):
    allFiles = list(os.listdir(dir))
    # print(allFiles)
    fileNameList = []
    numOfFiles = len(allFiles)
    i = 0
    for fichier in allFiles:
        if(i % getEveryNthLine == 0):
                # print(fichier)
                if(fichier.endswith(".csv")):
                    fileNameList.append(dir + "/" + fichier[0:-4])
        i += 1
    return fileNameList

print("Creating Model")
model = createModel()

# print("Model Summary")
# model.summary()

print("\nLoad Files")
files = getFile("F:/cartoonset10k/", 100)
print("Loaded " + str(len(files)) + " file names")

print("Split Data Into Train And Test")
train, test = train_test_split(files, test_size=0.2)

x_train = []
y_train = []
x_test = []
y_test = []

def getLabels(filePath):
    df = []
    with open(filePath, "r") as file:
        for line in list(file):
            tempList = line.replace("\n", "").replace('"', "").replace(" ", "").split(",")
            df.append({
                "attr": tempList[0],
                "value":int(tempList[1]),
                "maxValue":int(tempList[2])
            })
    return df

for i in range(len(train)):

    x_train.append([list(Image.open(train[i] + ".png").getdata())])
    y_train.append(pd.DataFrame(getLabels(train[i] + ".csv"))["value"][1])

print("Finished Formating\n\n")

x_train = np.reshape(x_train, (len(train), 500, 500, 4, 1))

with tf.device('/gpu'):
    model.fit(x_train, y_train, batch_size=8, epochs=3)

Does anyone know what is causing it? I think the problem has to do with the input shape, but I could be very wrong. Where can I find where the correct input shape should be?

Answer 1

A few problems there my friend:

First of all it is bad practice to have a kernal_size of even numbers, Instead i recommend something like (5, 5) since you image in considerably large

Next, I think you should be using Conv2D instead of Conv3D

Also, your final layer should have a few outputs like 2, representing maybe something like 0 for that cartoon character and 1 for not that cartoon character.

Finally, I keep getting a sense that your model will underfit and it is always good to overfit your model and combat it using strategies such as L1, L2 weight regularization or Droupout layers.

Overall your model should be something like this, feel free to call me off if I am wrong as I am no expert in this as well :P :

def build_model():
    model = Sequential()

    model.add(Conv2D(64, (5, 5), input_shape=(500, 500, 4), activation="relu"))
    model.add(MaxPooling2D(pool_size=(2, 2))
    model.add(Droupout(0.2))

    model.add(Conv2D(128, (5, 5), activation="relu"))
    model.add(MaxPooling2D(pool_size=(2, 2))
    model.add(Droupout(0.2))

    model.add(Conv2D(256, (5, 5), activation="relu"))
    model.add(MaxPooling2D(pool_size=(2, 2))
    model.add(Droupout(0.2))

    model.add(Flatten())
    model.add(Dense(512, activation="relu"))
    model.add(Dropout(0.5))
    model.add(Dense(512, activation="relu"))
    model.add(Dropout(0.5))
    model.add(Dense(512, activation="relu"))
    model.add(Dropout(0.5))
    model.add(Dropout(2, activation="softmax"))

return model

I used a few import statements so that I don't have to type k.blah.blah etc. CNN model usually goes: conv -> maxpool -> dropout -> conv -> maxpool -> dropout .......... dropout -> Flatten -> Dense -> High Dropout rate -> Dense -> High Dropout rate -> ........... -> output

Hope this helps! ciao

PS: You are only 13 years old and is able to code stuff like CNNs etc, I am really impressed, you have talents my friend. Keep up with the good work! :)