0

我正在创建一个 tensorflow 模型，该模型应该处理包含图像文件和其他分类输入的输入，两者都可以通过 csv 文件访问。首先我有这个模型，只有图像输入，但是当我用额外的元数据扩展输入时，我的model.fit.

这是我的代码：

import pandas as pd
import numpy as np
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers, regularizers
from keras.callbacks import History


df = pd.read_csv('dataset2.csv')
file_paths = df['file_name'].values 
labels = df['label'].to_numpy(dtype='float32')
metadata = df[['usage', 'completion', 'height', 'construction']].to_numpy(dtype='float32')


dataset_size = len(df.index)  
train_size = int(0.8 * dataset_size)
val_size = int(0.1 * dataset_size)
test_size = int(0.1 * dataset_size)


full_dataset_img = tf.data.Dataset.from_tensor_slices(file_paths)
ds_train_img = full_dataset_img.take(train_size)
ds_test_img = full_dataset_img.skip(train_size)
ds_val_img = ds_test_img.skip(val_size)
ds_test_img = ds_test_img.take(test_size)

labels = np.reshape(labels, (dataset_size, 1))
full_dataset_lbl = tf.data.Dataset.from_tensor_slices(labels)
ds_train_lbl = full_dataset_lbl.take(train_size)
ds_test_lbl = full_dataset_lbl.skip(train_size)
ds_val_lbl = ds_test_lbl.skip(val_size)
ds_test_lbl = ds_test_lbl.take(test_size)

metadata = np.reshape(metadata, (dataset_size, 4))
full_dataset_dat = tf.data.Dataset.from_tensor_slices(metadata)
ds_train_dat = full_dataset_dat.take(train_size)
ds_test_dat = full_dataset_dat.skip(train_size)
ds_val_dat = ds_test_dat.skip(val_size)
ds_test_dat = ds_test_dat.take(test_size)



# FUNCTION TO READ AND NORMALIZE THE IMAGES AND METADATA
def read_image(image_file):
    image = tf.io.read_file(image_file) 
    image = tf.image.decode_jpeg(image, channels=3)
    image = tf.image.resize(image, (350, 350)) 
    return tf.cast(image, tf.float32) / 255.0


# FUNCTION FOR DATA AUGMENTATION
def augment(image):
    if tf.random.uniform((), minval=0, maxval=1) < 0.1:
        image = tf.tile(tf.image.rgb_to_grayscale(image), [1, 1, 3])   
    image = tf.image.random_brightness(image, max_delta=0.25)
    image = tf.image.random_contrast(image, lower=0.75, upper=1.25)
    image = tf.image.random_saturation(image, lower=0.75, upper=1.25)
    image = tf.image.random_flip_left_right(image)  
    return image



ds_train_img = ds_train_img.map(read_image)
ds_train_img = ds_train_img.map(augment)
ds_val_img = ds_val_img.map(read_image)
ds_test_img = ds_test_img.map(read_image)

ds_train_lbl = np.array(ds_train_lbl)
ds_train_dat = np.array(ds_train_dat)



# DEFINING FUNCTIONAL MODEL FOR COMBINING IMAGE AND METADATA INPUTS
input_img = keras.Input(shape=(350, 350, 3))
input_dat = keras.Input(shape=(4,))

x = layers.Conv2D(16, (3, 3), activation='relu', kernel_regularizer=regularizers.l2(0.02), padding='same')(input_img)
x = layers.BatchNormalization(momentum=0.9)(x)
x = layers.MaxPooling2D()(x)
x = layers.Conv2D(32, (3, 3), activation='relu', kernel_regularizer=regularizers.l2(0.02), padding='same')(x)
x = layers.BatchNormalization(momentum=0.9)(x)
x = layers.MaxPooling2D()(x)
x = layers.Conv2D(64, (3, 3), activation='relu', kernel_regularizer=regularizers.l2(0.02), padding='same')(x)
x = layers.BatchNormalization(momentum=0.9)(x)
x = layers.MaxPooling2D()(x)
x = layers.Conv2D(128, (3, 3), activation='relu', kernel_regularizer=regularizers.l2(0.02), padding='same')(x)
x = layers.BatchNormalization(momentum=0.9)(x)
x = layers.MaxPooling2D()(x)
x = layers.Flatten()(x)
x = layers.Flatten()(x)

data = layers.Dense(128, activation='relu')(input_dat)

multi = layers.concatenate([x, data])
multi = layers.Dense(256, activation='relu')(multi)
multi = layers.Dropout(0.35)(multi)
output = layers.Dense(8, activation='sigmoid')(multi)
model = keras.Model(inputs=[input_img, input_dat], outputs=output)


model.compile(
    optimizer=keras.optimizers.Adam(3e-5),     
    loss=[keras.losses.SparseCategoricalCrossentropy()],
    metrics=["accuracy"])


model.fit(x=[ds_train_img, ds_train_dat], y=ds_train_lbl,
          validation_data=([ds_val_img, ds_val_dat], ds_val_lbl),
          epochs=30, batch_size=16, verbose=1)

当我想运行代码时，在 model.fit 中出现以下错误：

ValueError: Failed to find data adapter that can handle input: (<class 'list'> containing values of types {"<class 'tensorflow.python.data.ops.dataset_ops.MapDataset'>", "<class 'numpy.ndarray'>"}), <class 'numpy.ndarray'>

任何帮助是极大的赞赏！

4

0 回答 0