How to get stride value in Conv2D layer Tensorflow?

Question

My project is visualization of CNN 2d model and I have to get a NN-layer object from checkpoint model file. Does Tensorflow allow user to get layer object?

In fact, I just want the stride or padding value from Conv2d or Relu.

Answer 1

It's hard to get a Layer object from the checkpoint only. But you can get the convolutional attributes programmatically by calling:

tf.get_default_graph().as_graph_def()

... assuming that you've restored the checkpoint into a default graph. It returns the list of graph nodes (GraphDef proto to be exact), among which there are also Conv2D nodes, like this one:

node {
  name: "conv2d/Conv2D"
  op: "Conv2D"
  input: "input_layer"
  input: "conv2d/kernel/read"
  attr {
    key: "T"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "data_format"
    value {
      s: "NHWC"
    }
  }
  attr {
    key: "padding"
    value {
      s: "SAME"
    }
  }
  attr {
    key: "strides"
    value {
      list {
        i: 1
        i: 1
        i: 1
        i: 1
      }
    }
  }
  attr {
    key: "use_cudnn_on_gpu"
    value {
      b: true
    }
  }
}

You just need to find the node that corresponds to the particular layer you're interested in.