Tensorflow transfer learning fine tunning 改进图像训练结果实践

改进图像训练结果的常用方法是通过随机方式变形，裁剪或增亮训练输入

–random_crop

–random_scale

–random_brightness

传递给脚本来启用这些失真。这些都是控制每个图像应用多少失真的百分比值。对于每个值开始使用5或10的值是合理的，然后通过实验看看哪些值有助于的应用程序。

–flip_left_right将水平随机地镜像一半的图像，只要那些颠倒可能发生在你的应用程序。

–random_brightness 5

/media/haijunz/27a263b4-e313-4b58-a422-0201e4cb11ed/tensorflow/tensorflow/tensorflow/examples/image_retraining$ python retrain.py –image_dir /media/haijunz/27a263b4-e313-4b58-a422-0201e4cb11ed/tensorflow/models/flower_photos/data –random_brightness 5

W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn’t compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.

W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn’t compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.

W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn’t compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.

Looking for images in ‘roses’

Looking for images in ‘sunflowers’

Looking for images in ‘tulips’

Looking for images in ‘dandelion’

Looking for images in ‘daisy’

W tensorflow/core/framework/op_def_util.cc:332] Op BatchNormWithGlobalNormalization is deprecated. It will cease to work in GraphDef version 9. Use tf.nn.batch_normalization().

2017-03-14 15:08:42.443384: Step 0: Train accuracy = 69.0%

2017-03-14 15:08:42.443478: Step 0: Cross entropy = 1.527363

2017-03-14 15:08:42.596276: Step 0: Validation accuracy = 60.0% (N=100)

……