search

forresti / SqueezeNet

SqueezeNet: AlexNet-level accuracy with 50x fewer parameters
BSD 2-Clause "Simplified" License
2.17k stars 723 forks source link

conv10 layer has pad 1 #32

Closed psyhtest closed 7 years ago

psyhtest commented 7 years ago

As noticed by @milakov, it's quite surprising to see the following 1x1 convolutional layer having the padding of 1:.

layer {
  name: "conv10"
  type: "Convolution"
  bottom: "fire9/concat"
  top: "conv10"
  convolution_param {
    num_output: 1000
    pad: 1
    kernel_size: 1
  }
}

Any comments?

forresti commented 7 years ago

Duplicate issue. See #14

psyhtest commented 7 years ago

@forresti Thanks! (I really should have guessed this as a compiler person used to generating correct if redundant code :))

forresti commented 7 years ago

Haha yup, that's exactly what happened. :)