0

我按照 tensorflow 网站上关于数据集创建的教程编写了以下图像分类代码。问题是在创建数据集变量后，代码只是停止执行并终止。

path_ds = tf.data.Dataset.from_sparse_tensor_slices(all_images)
# PROGRAM TERMINATES AND DOES NOT EXECUTE THE FOLLOWING LINE 
image_ds = path_ds.map(load_image)

我最初的猜测是，这可能是急切执行的问题。我尝试使用 Visual Studio 代码调试器进行调试，但无法取得任何进展。

import os
import numpy as np 
import tensorflow as tf 
from PIL import Image

tf.compat.v1.enable_eager_execution()

test_folder = "../Test"
train_folder = "../Train"
BATCH_SIZE = 32

folders = os.listdir(train_folder)
labels = sorted(folders, key = lambda x : int(x.split("_")[1]) if "c" in x else \
                                  int(x.split("_")[1])+ 100)
char_to_int = dict((label, index) for index, label in enumerate(labels))
int_to_char = dict((index, label) for index, label in enumerate(labels))

def load_image(infilename) :
    img = Image.open( infilename )
    img.load()
    data = np.asarray( img, dtype="int32" )
    return tf.convert_to_tensor(np.expand_dims(data, axis=2))

def get_all_image_names(folder):
    children = os.listdir(folder)
    all_images = []
    all_labels = []
    for i in children:
        images = os.listdir(os.path.join(folder,i))
        for image in images:
            path = os.path.join(folder, i, image)
            all_images.append(path)
            all_labels.append(char_to_int[i])
    path_ds = tf.data.Dataset.from_sparse_tensor_slices(all_images)


    # PROGRAM TERMINATES AFTER THIS LINE 

    image_ds = path_ds.map(load_image)
    label_ds = tf.data.Dataset.from_sparse_tensor_slices(tf.cast(all_labels, tf.int64))
    image_label_ds = tf.data.Dataset.zip((image_ds, label_ds))
    return image_label_ds

dataset = get_all_image_names(train_folder)

ds = dataset.shuffle(buffer_size=10000)
ds = ds.batch(BATCH_SIZE)

model = tf.keras.models.Sequential([
  tf.keras.layers.Flatten(input_shape=(32, 32,1)),
  tf.keras.layers.Dense(512, activation=tf.nn.relu),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(64, activation=tf.nn.relu),
  tf.keras.layers.Dropout(0.2),
  tf.keras.layers.Dense(46, activation=tf.nn.softmax)
])

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

model.fit(ds, epochs=55, verbose=1)

print("MOdel fitted")
4

0 回答 0