Skip to content

concept y_pred[:, 2:, :] and downsample_factor ?  #21

Description

@PythonImageDeveloper

Hi
Q1 - I don't understand why don't use all of y_pred[ :,:,:]? why use y_pred[ :, 2:,:]? why don't use 0,1 dims?

def ctc_lambda_func(args):
    y_pred, labels, input_length, label_length = args
    # the 2 is critical here since the first couple outputs of the RNN
    # tend to be garbage:
    y_pred = y_pred[:, 2:, :]
    return K.ctc_batch_cost(labels, y_pred, input_length, label_length)

Q2 - What's meaning the downsampling parameter? if we change the input_size, is necessary to change the value of downsampling parameter? what's the principle of the parameter?
and I also don't understand, why use this line and why we multiply (self.img_w // self.downsample_factor - 2) to np.ones((self.batch_size, 1)) ? what's the advantage of this work? if we use only np.ones((self.batch_size, 1)) , Causing the problem?
and why use (self.downsample_factor - 2) ? why 2 ?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions