Pooling layer output for cifar10_quick_train_test is not an integer

Asked 2 years ago, Updated 2 years ago, 106 views

Caffe cifar10_quick_train_test has the output of the pooling layer

OH=(H-FH)/S+1
(32-3)/2 + 1
29/2 + 1

The calculations are as follows:
I don't think the number will be divisible, but what kind of process is being done?

https://github.com/BVLC/caffe/blob/master/examples/cifar10/cifar10_quick_train_test.prototxt

layer{
  name: "conv1"
  type: "Convolution"
  bottom: "data"
  top: "conv1"
  param{
    lr_mult: 1
  }
  param{
    lr_mult:2
  }
  convolution_param{
    num_output —32
    pad:2
    kernel_size —5
    stride:1
    weight_filler{
      type: "gaussian"
      std —0.0001
    }
    bias_filler{
      type: "constant"
    }
  }
}
layer{
  name: "pool1"
  type: "Pooling"
  bottom: "conv1"
  top: "pool1"
  pooling_param{
    pool —MAX
    kernel_size:3
    stride —2
  }
}

python machine-learning deep-learning caffe

2022-09-30 17:32

1 Answers

If not, the response depends on the deep learning framework (software) and some output errors or round to the nearest integer.

https://github.com/BVLC/caffe/blob/128797ebe100dd87f375d2d30b0f649914d0fcc3/src/caffe/layers/pooling_layer.cpp#L91

Looking around here, caffe allows you to change the operating mode with parameters.


2022-09-30 17:32

If you have any answers or tips


© 2024 OneMinuteCode. All rights reserved.