Global max pooling operation for spatial data.
layer_global_max_pooling_2d(object, data_format = NULL, keepdims = FALSE, ...)
What to call the new
A string, one of
A boolean, whether to keep the spatial dimensions or not. If
standard layer arguments.
data_format='channels_last': 4D tensor with shape:
(batch_size, rows, cols, channels)
data_format='channels_first': 4D tensor with shape:
(batch_size, channels, rows, cols)
2D tensor with shape:
Other pooling layers: