qq_34500588的博客

tfrecords文件过大问题使用默认方式将图像转成tfrecords文件很大使用默认方式将图像转成tfrecords文件很大解决方法：构造writer时，加上options参数——options = tf.python_io.TFRecordOptions(compression_type=tf.python_io.TFRecordCompressionType.ZLIB)writer = tf.python_io.TFRecordWriter(tfrecords_path, options)

参考TensorFlow官方文档TF官方文档中文

meta转化为summary用tensorboard查看import tensorflow as tffrom tensorflow.python.platform import gfilegraph = tf.get_default_graph()graphdef = graph.as_graph_def()_ = tf.train.import_meta_graph('meta的路...

slim代码示例，包括读取tfrecords，带BN的训练，L2正则化读取tfrecordsimport globimport cv2import numpy as npimport tensorflow as tf# 从自定义的config.py里导入一些参数from config import HEIGHT, WIDTH, CHANNELdef read(file_queue...

TensorFlow用GPU训练#配置GPUconfig = tf.ConfigProto()config.gpu_options.per_process_gpu_memory_fraction = 0.7 # 显存占用率config.allow_soft_placement = Trueconfig.log_device_placement = Trueconfig.gpu_opt...

TensorFlow模型变量重用问题加载模型及检测的.py# predict.pyimport numpy as npimport tensorflow as tffrom PIL import Imageimport timeimport fcrndef predict(model_data_path, image_path , png_path): # Defau...

TensorFlow slim中正则化的使用关于slim中的正则化使用，以L2正则化为例：# 构建网络时候添加l2正则化with slim.arg_scope([slim.conv2d], weights_regularizer=slim.l2_regularizer(0.0001)): net = slim.conv2d(inputs, 64, 5, 2, padding='SAME',...

TensorFlow实现batch_bormdef batch_norm(x, is_training, eps=EPS, decay=0.9, affine=True, name='batch_norm'): from tensorflow.python.training.moving_averages import assign_moving_average ...

TensorFlow实现2×2反池化2×2反池化如下图，将输入的尺寸扩大为原来两倍，输入值填充到新的每个2×2网格的左上角，其余三个填0。代码实现TensorFlow中没有反池化函数，以下是代码实现。# 2x2反池化def unpool(value, name='unpool'): with tf.name_scope(name) as scope: sh =...