无法广播阵列

Question

我正在读取图像数据以构建分类器。在它完美运行之前，我只给了我一半的数据（50-60％），但是如果我给我所有的数据（100％），它给了我一个混乱的错误。我所有的图像都是彩色的（？，？，3）我已经检查过了。但是仍然给了我形状错误。请帮忙。

代码

from sklearn.preprocessing import OneHotEncoder
import tensorflow as tf
import numpy as np
import scipy.io as cio
import os
import matplotlib.pyplot as plt
import matplotlib.image as mpg
from random import shuffle
import tflearn
from tflearn.layers.conv import conv_3d, max_pool_3d
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.estimator import regression
import cv2

a = cio.loadmat("D:/compCarsThesisData/data/misc/make_model_name.mat")

images = "D:\compCarsThesisData\data\image"
# images = "D:/practiceData/data/image/"
IMG_SIZE = 64
MODEL_NAME = 'Classification'
LR = 1e-3

b = a['make_names']
# c = b.reshape(163,)
d = []
count = 0
for i in range(b.size):
  d.append(b[i][0][0])
  #print(d)

labels_dic = {v: k for v, k in enumerate(d)}
print(labels_dic)

image1 = []
label1 = []

for root, _, files in os.walk(images):
  cdp = os.path.abspath(root)

  for f in files:

    name,ext = os.path.splitext(f)
    boolean = True

    if ext == ".jpg":
      cip = os.path.join(cdp,f)      
      ci = mpg.imread(cip)
      image = cv2.cv2.resize(ci,(IMG_SIZE,IMG_SIZE))
      np.array(image1.append(image))

      arr = np.zeros(163)
      print(cip.split('\\')[4])
      arr[int(cip.split('\\')[4])-1] = 1

      # exit()
      label1.append(arr)
      count = count+1
      print(count)

print("done :)")


training_data = []

training_data.append([np.array(image1),np.array(label1)])
#print("TrainingData",training_data)
shuffle(training_data)
# np.save('training_data_with_One_Hot', training_data)

testing_data = []
testing_data.append([np.array(image1),np.array(label1)])
#print("TestingDATA",testing_data)
# np.save('testing_data_with_One_Hot',testing_data)
shuffle(testing_data)

# training_data = np.load('training_data_with_One_Hot.npy')
# testing_data = np.load('testing_data_with_One_Hot.npy')

# print("Training Data Shape",training_data)
print(testing_data[0][1].shape)
train = training_data[-2000:]
test = testing_data[-2000:]

print("TRAINS",train[0][0].shape)
print("TEST SHAPE",test[0][0][:1].shape)
X_train = train[0][0] #np.array(train[0]).reshape(None, IMG_SIZE, IMG_SIZE, 3)

print("XTRAINING DATA",X_train.shape)
y_train = train[0][1]#np.array([i[0][1] for i in train])
print("Ytrain",y_train.shape)
X_test = test[0][0]
print("XTEST", X_test.shape) #np.array([i[0] for i in test]).reshape(None, IMG_SIZE, IMG_SIZE, 3)
y_test = test[0][1] #np.array([i[1] for i in test])
print("YTEST",y_test.shape)

# sess = tf.Session()cls


tf.reset_default_graph()
convnet = input_data(shape=(None,IMG_SIZE,IMG_SIZE,3),name='input')
#shape=[None, IMG_SIZE, IMG_SIZE, 1],
convnet = conv_3d(convnet, 32, 5, activation='relu')
convnet = max_pool_3d(convnet, 5)
convnet = conv_3d(convnet, 64, 5, activation='relu')
convnet = max_pool_3d(convnet, 5)
convnet = conv_3d(convnet, 128, 5, activation='relu')
convnet = max_pool_3d(convnet, 5)
convnet = conv_3d(convnet, 64, 5, activation='relu')
convnet = max_pool_3d(convnet, 5)
convnet = conv_3d(convnet, 32, 5, activation='relu')
convnet = max_pool_3d(convnet, 5)
convnet = fully_connected(convnet, 1024, activation='relu')
convnet = dropout(convnet, 0.8)
convnet = fully_connected(convnet, 163, activation='softmax')
convnet = regression(convnet, optimizer='adam', learning_rate=LR, loss='categorical_crossentropy', name='targets')
model = tflearn.DNN(convnet, tensorboard_dir='log', tensorboard_verbose=0)
model.fit({'input': X_train}, {'targets': y_train}, n_epoch=10,
          validation_set=({'input': X_test}, {'targets': y_test}),
          snapshot_step=500, show_metric=True, run_id=MODEL_NAME)

错误

Traceback (most recent call last):
  File "d:/ThesisWork/seriouswork/classifier_with_onehot.py", line 83, in <module>
    training_data.append([np.array(image1),np.array(label1)])
ValueError: could not broadcast input array from shape (64,64,3) into shape (64,64)

错误在第83行上，它是training_data.append，我不知道我在这里实际上在做什么错。为什么在我提供一些数据但添加更多数据时会出错，为什么它起作用。 另外，我正在根据文件名生成自己的标签，例如一种热编码的文件，以便稍后对我的数据进行分类，这就是为什么arr是np.zeros（163）的原因，因为我在163个不同的文件夹中有数据。例如，D：/ compCarsThesisData / data / image / 1 / ...， D：/ compCarsThesisData / data / image / 2 / ... D：/ compCarsThesisData / data / image / 3 / ...等等，直到D：/ compCarsThesisData / data / image / 163 /...