98 lines
3.9 KiB
Python
98 lines
3.9 KiB
Python
import tensorflow as tf
|
|
import os
|
|
|
|
class read_Input_Data:
|
|
def __init__(self) -> None:
|
|
pass
|
|
def save_tfrecords(self, images, label):
|
|
'''將資料儲存為TFRecord數據'''
|
|
image_width, image_height = 64, 64
|
|
image_channel = 3
|
|
tfrecod_data_root = "../../Dataset/tfrecode_Dataset/tfrecod_data.tfrecords"
|
|
if not os.path.exists(tfrecod_data_root):
|
|
os.makedirs(tfrecod_data_root)
|
|
|
|
TFWriter = tf.python_io.TFRecordWriter(tfrecod_data_root)
|
|
|
|
try:
|
|
for i in range(len(images)):
|
|
if images[i] is None:
|
|
print('Error image:' + images[i])
|
|
else:
|
|
#圖片轉為字串
|
|
image_raw = str(images[i])
|
|
|
|
|
|
# 將 tf.train.Feature 合併成 tf.train.Features
|
|
train_feature = tf.train.Features(feature={
|
|
'Label' : self.int64_feature(label),
|
|
'image_raw' : self.bytes_feature(image_raw),
|
|
'channel' : self.int64_feature(image_channel),
|
|
'width' : self.int64_feature(image_width),
|
|
'height' : self.int64_feature(image_height)}
|
|
)
|
|
|
|
# 將 tf.train.Features 轉成 tf.train.Example
|
|
train_example = tf.train.Example(features = train_feature)
|
|
|
|
# 將 tf.train.Example 寫成 tfRecord 格式
|
|
TFWriter.write(train_example.SerializeToString())
|
|
|
|
except Exception as e:
|
|
print(e)
|
|
|
|
TFWriter.close()
|
|
print('Transform done!')
|
|
|
|
return tfrecod_data_root
|
|
|
|
# 轉Bytes資料為 tf.train.Feature 格式
|
|
def int64_feature(self, value):
|
|
if not isinstance(value, list):
|
|
value = [value]
|
|
return tf.train.Feature(int64_list=tf.train.Int64List(value=value))
|
|
|
|
def bytes_feature(self, value):
|
|
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
|
|
|
|
def Data_Decompile(self, example):
|
|
'''反編譯TFR數據'''
|
|
feature_description = {
|
|
'data': tf.io.FixedLenFeature([], tf.string),
|
|
'label': tf.io.FixedLenFeature([], tf.float32),
|
|
}
|
|
parsed_example = tf.io.parse_single_example(example, features=feature_description)
|
|
|
|
x_sample = tf.io.parse_tensor(parsed_example['data'], tf.float32)
|
|
y_sample = parsed_example['label']
|
|
|
|
return x_sample, y_sample
|
|
|
|
def load_dataset(self, filepaths):
|
|
'''
|
|
載入TFR數據集
|
|
* dataset.shuffle(shuffle_buffer_size):
|
|
隨機打亂此數據集的元素。
|
|
|
|
該數據集用 buffer_size 元素填充緩衝區，然後從該緩衝區中隨機採樣元素，用新元素替換所選元素。
|
|
為了完美改組，需要緩衝區大小大於或等於數據集的完整大小。
|
|
|
|
例如，如果您的數據集包含 10,000 個元素，但 buffer_size 設置為 1,000,
|
|
則 shuffle 最初只會從緩衝區的前 1,000 個元素中選擇一個隨機元素。
|
|
一旦選擇了一個元素，它在緩衝區中的空間將被下一個（即第 1,001 個）元素替換，從而保持 1,000 個元素的緩衝區。
|
|
'''
|
|
shuffle_buffer_size = 700
|
|
batch_size = 128
|
|
tfrecod_data_root = "../../Dataset/tfrecode_Dataset"
|
|
|
|
dataset = tf.data.TFRecordDataset(filepaths)
|
|
dataset = dataset.shuffle(shuffle_buffer_size)
|
|
dataset = dataset.map(map_func=self.Data_Decompile, num_parallel_calls= 8)
|
|
dataset = dataset.batch(batch_size).prefetch(64)
|
|
|
|
# 產生文件名隊列
|
|
filename_queue = tf.train.string_input_producer([filename],
|
|
shuffle=True,
|
|
num_epochs=3)
|
|
|
|
return dataset |