MemoryDataLayer label data dimension

329 views

Skip to first unread message

Mark Laagland

unread,

Mar 23, 2015, 11:36:14 AM3/23/15

to caffe...@googlegroups.com

Hi all,

First, let me admit I do not know nearly as much about deep learning as I should, as a student, I am still learning. If I say something that makes no sense, pointers would be greatly appreciated.

I am trying to use Caffe to build a CNN for facial keypoint extraction (the Kaggle competition). I started with this tutorial which, with a lot of reading, trying and cursing (i.e. learning), I almost got it working. Almost, because the network outputs the same data for any input. To figure out why, I started rewriting the code to use the pyCaffe interface. (Has always been the goal and should provide more information for troubleshooting).

The data consists of images (96x96 greyscale pixels) for which 15 keypoints need to be found.
So, the label data consists of 30 values representing the 15 keypoint coordinates.

Unfortunately, I cannot figure out how to set up the MemoryDataLayer properly to accept the label data. I keep on getting errors like the following out of my network.

euclidean_loss_layer.cpp:14] Check failed: bottom[0]->count(1) == bottom[1]->count(1) (30 vs. 1) Inputs must have the same dimension.

Obviously, I need to set the dimensions of the label data somewhere. But how? By now I am quite sure I cannot set this up in the layer parameters itself and am out of leads.

Some pointers would be very much appreciated.

N.B. I think my final code could be a nice addition to the Notebook Examples. I'd be more than willing to share it.

My full network setup (currently mostly the same as the tutorial):

name: "VER1"

layers {
  name: "i_data"
  top: "data"
  top: "label"
  type: MEMORY_DATA
  memory_data_param {
    batch_size: 64
    channels: 1
    height: 96
    width: 96
  }
}

layers {
  name: "conv1"
  type: CONVOLUTION
  bottom: "data"
  top: "conv1"
  convolution_param {
    num_output: 32
    kernel_size: 11
    stride: 2
   
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu2"
  type: RELU
  bottom: "conv1"
  top: "conv1"
}

layers {
  name: "pool1"
  type: POOLING
  bottom: "conv1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}

layers {
  name: "conv2"
  type: CONVOLUTION
  bottom: "pool1"
  top: "conv2"
  convolution_param {
    num_output: 64
    pad: 2
    kernel_size: 7
    group: 2
   
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu2"
  type: RELU
  bottom: "conv2"
  top: "conv2"
}

layers {
  name: "pool2"
  type: POOLING
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}

layers {
  name: "norm2"
  type: LRN
  bottom: "pool2"
  top: "norm2"
  lrn_param {
    norm_region: WITHIN_CHANNEL
    local_size: 3
    alpha: 5e-05
    beta: 0.75
  }
}

layers {
  name: "conv3"
  type: CONVOLUTION
  bottom: "norm2"
  top: "conv3"
  convolution_param {
    num_output: 32
    pad: 1
    kernel_size: 5
    
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu3"
  type: RELU
  bottom: "conv3"
  top: "conv3"
}

layers {
  name: "conv4"
  type: CONVOLUTION
  bottom: "conv3"
  top: "conv4"
  convolution_param {
    num_output: 64
    pad: 1
    kernel_size: 5
    
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu4"
  type: RELU
  bottom: "conv4"
  top: "conv4"
}

layers {
  name: "conv5"
  type: CONVOLUTION
  bottom: "conv4"
  top: "conv5"
  convolution_param {
    num_output: 32
    pad: 1
    kernel_size: 5
    
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu5"
  type: RELU
  bottom: "conv5"
  top: "conv5"
}

layers {
  name: "pool5"
  type: POOLING
  bottom: "conv5"
  top: "pool5"
  pooling_param {
    pool: MAX
    kernel_size: 4
    stride: 2
  }
}

layers {
  name: "drop0"
  type: DROPOUT
  bottom: "pool5"
  top: "pool5"
  dropout_param {
    dropout_ratio: 0.5
  }
}
layers {
  name: "ip1"
  type: INNER_PRODUCT
  bottom: "pool5"
  top: "ip1"
  inner_product_param {
    num_output: 100
    
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu4"
  type: RELU
  bottom: "ip1"
  top: "ip1"
}

layers {
  name: "drop1"
  type: DROPOUT
  bottom: "ip1"
  top: "ip1"
  dropout_param {
    dropout_ratio: 0.5
  }
}

layers {
  name: "ip2"
  type: INNER_PRODUCT
  bottom: "ip1"
  top: "ip2"
  inner_product_param {
    num_output: 30
    
    bias_filler {
      type: "constant"
      value: 0.1
    }
  }
}

layers {
  name: "relu22"
  type: RELU
  bottom: "ip2"
  top: "ip2"
}

layers {
  name: "loss"
  type: EUCLIDEAN_LOSS
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}

Siddhant Ranade

unread,

May 16, 2016, 4:07:46 AM5/16/16

to Caffe Users

Were you able to figure this out? I am also stuck here. I am currently using HDF5 to avoid this issue.