vgg16的tensorflow网络代码

[double-vane]

你好，vgg16的tensorflow网络代码在哪里？你的deploy.prototxt里网络层的名称是随便取的吗？

[abner2015]

tf2caffe.py 里面，网络名称对应就可以

[double-vane]

额，我的意思是您这里的vgg16.ckpt的链接已经失效，所以我看不到里面网络的名称。 我说的网络代码是指构造vgg16的网络代码，是用slim构造的，在tensorflow里的/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/slim/python/slim/nets/vgg.py构造的吗？ 您能帮我看一下，如果我的网络显示是这样的，那我写.prototxt时，第一层的conv的名称是随便取还是local/Conv?十分感谢 Converted 28 variables to const ops. 102 ops in the final graph. (u'Placeholder', (<tf.Tensor 'Placeholder:0' shape=(32, 224, 224, 3) dtype=float32>,)) (u'Placeholder_1', (<tf.Tensor 'Placeholder_1:0' shape=(32, 8) dtype=float32>,)) (u'Placeholder_2', (<tf.Tensor 'Placeholder_2:0' shape=(32, 224, 224, 3) dtype=float32>,)) (u'Placeholder_3', (<tf.Tensor 'Placeholder_3:0' shape=(32, 8) dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform/shape', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/shape:0' shape=(4,) dtype=int32>,)) (u'local/Conv/weights/Initializer/random_uniform/min', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/min:0' shape=() dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform/max', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/max:0' shape=() dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform/RandomUniform', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/RandomUniform:0' shape=(3, 3, 3, 16) dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform/sub', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/sub:0' shape=() dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform/mul', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform/mul:0' shape=(3, 3, 3, 16) dtype=float32>,)) (u'local/Conv/weights/Initializer/random_uniform', (<tf.Tensor 'local/Conv/weights/Initializer/random_uniform:0' shape=(3, 3, 3, 16) dtype=float32>,)) (u'local/Conv/weights', (<tf.Tensor 'local/Conv/weights:0' shape=(3, 3, 3, 16) dtype=float32_ref>,)) (u'local/Conv/weights/Assign', (<tf.Tensor 'local/Conv/weights/Assign:0' shape=(3, 3, 3, 16) dtype=float32_ref>,)) (u'local/Conv/weights/read', (<tf.Tensor 'local/Conv/weights/read:0' shape=(3, 3, 3, 16) dtype=float32>,)) (u'local/Conv/kernel/Regularizer/l2_regularizer/scale', (<tf.Tensor 'local/Conv/kernel/Regularizer/l2_regularizer/scale:0' shape=() dtype=float32>,)) (u'local/Conv/kernel/Regularizer/l2_regularizer/L2Loss', (<tf.Tensor 'local/Conv/kernel/Regularizer/l2_regularizer/L2Loss:0' shape=() dtype=float32>,)) (u'local/Conv/kernel/Regularizer/l2_regularizer', (<tf.Tensor 'local/Conv/kernel/Regularizer/l2_regularizer:0' shape=() dtype=float32>,))

[abner2015]

我这个是做模型转换的，首先是需要有训练好的模型，也就是网络结构是定义好的，所以网络可以根据自己需要来定义，定义完训练。如果是有模型文件，那不能随便定义