How to predict a single digit using Caffe and C++

[Yaroslav Melekhov]

 I am trying to use Caffe to predict a digit in MNIST dataset. I run a solver to pretrain a net as follow:

    // Set to TRAIN Phase
    caffe::Caffe::set_phase(caffe::Caffe::TRAIN);
    // set GPU
    caffe::Caffe::set_mode(caffe::Caffe::GPU);
    int deviceId = 0;
    caffe::Caffe::SetDevice(deviceId);
    LOG(INFO) << "Using GPU";
    // Solver initialization and evaluation
    caffe::SolverParameter solverParams;
    caffe::ReadProtoFromTextFileOrDie("/home/iar/caffe/caffe/examples/mnist/lenet_solver.prototxt", &solverParams);
    caffe::Solver<float>* solver = new caffe::SGDSolver<float>(solverParams);
    solver->Solve();

I just use LeNet prototxt file from examples to do this. Ok, this stage is crystal clear. Then I found a script which converts mnist dataset (lmdb) to a set of grayscale images (.jpg) each of them is a representation of a handwritten digit. I added Image-Data layer as mention here (). Therefore, my .prototxt file is (my_lenet_test.prototxt):

name: "LeNet"
layers {
  name: "mnist"
  type: DATA
  top: "data"
  top: "label"
  data_param {
    source: "/home/iar/caffe/caffe/examples/mnist/mnist_train_lmdb"
    backend: LMDB
    batch_size: 64
  }
  transform_param {
    scale: 0.00390625
  }
  include: { phase: TRAIN }
}
layers {
  name: "data"
  type: IMAGE_DATA
  top: "data"
  top: "label"
  image_data_param {
    source: "/home/iar/caffe/caffe/examples/mnist/file_list.txt"
    batch_size: 1
    new_height: 28
    new_width: 28
  }
  include: { phase: TEST }
}

layers {
  name: "conv1"
  type: CONVOLUTION
  bottom: "data"
  top: "conv1"
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 20
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
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"
  blobs_lr: 1
  blobs_lr: 2
  convolution_param {
    num_output: 50
    kernel_size: 5
    stride: 1
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "pool2"
  type: POOLING
  bottom: "conv2"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layers {
  name: "ip1"
  type: INNER_PRODUCT
  bottom: "pool2"
  top: "ip1"
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 500
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "relu1"
  type: RELU
  bottom: "ip1"
  top: "ip1"
}
layers {
  name: "ip2"
  type: INNER_PRODUCT
  bottom: "ip1"
  top: "ip2"
  blobs_lr: 1
  blobs_lr: 2
  inner_product_param {
    num_output: 10
    weight_filler {
      type: "xavier"
    }
    bias_filler {
      type: "constant"
    }
  }
}
layers {
  name: "accuracy"
  type: ACCURACY
  bottom: "ip2"
  bottom: "label"
  top: "accuracy"
  include: { phase: TEST }
}
layers {
  name: "loss"
  type: SOFTMAX_LOSS
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}
layers {
  name: "output"
  type: ARGMAX
  bottom: "loss"
  top: "output"
  include: { phase: TEST }
}

Then I construct my net:

caffe::Net<float> mnistLeNet("my_lenet_test.prototxt");

and load pre-trained layers from the solver

mnistLeNet.CopyTrainedLayersFrom("/home/iar/caffe/caffe/examples/mnist/lenet_iter_10000.caffemodel");

But on the last stage I got SIGABRT signal from my IDE and got the error:

Check failed: target_blobs[j]->channels() == source_layer.blobs(j).channels() (3 vs. 1) 

I  would highly appreciate for any help. I am stuck and have no idea what to do... Thank you!

[Friendly User]

hint: greyscale vs color (3 channels)

[Yaroslav Melekhov]

Yes, I understand the difference between grayscale and color images but do not understand what do you mean :). Actually, what target_blobs and source_layer.blobs() are? I think that source_layer.blobs() is my input image but target_blobs?

[sunil surineni]

  include: { phue: TEST }
}

Hei did you come with any solution . why dont you check out

... this http://ibuzzlog.blogspot.tw/search/label/Caffe

please let me know if you found out 

[J. Yegerlehner]

It's telling you that the net expects a blob with a shape that has 1 color channel. That's what you created when you trained against the gray scale MNIST digits. When you load the jpg images, those have 3 channels. One for each of Red, green and blue. They don't match. You can't create a network that has inputs of a certain shape, and then give it an input that has the wrong shape.

[Pasta farianist]

So how would I tell Caffe to either load images as grayscale or drop two out of three channels?

суббота, 28 февраля 2015 г., 1:48:06 UTC+3 пользователь J. Yegerlehner написал:

[Pastafarianist]

Well, I have found a solution, but it only works in very recent versions (initialneil's build for Windows is too old). You have to specify is_color: false in IMAGE_DATA layer, like this:

layers {

  name: "mnist"

  type: IMAGE_DATA

  top: "data"

  top: "label"

  image_data_param {

    source: "mnist_train_index.txt"

    batch_size: 64

    is_color: false

  }

  transform_param {

    scale: 0.00390625

  }

  include: { phase: TRAIN }

}

понедельник, 9 марта 2015 г., 20:18:25 UTC+3 пользователь Pastafarianist написал:

[Michael George Hart]

I have been getting this error
      F0411 20:48:10.789644 28542 blob.cpp:454] Check failed: ShapeEquals(proto) shape mismatch (reshape not set)
And believed it is/was the same problem that you are/were having....

I pulled the caffe.master on 03/31/2015
and still have the same issue.
My my network AlexNet trained with gray-scale images
Do you have any other recommendations?
Where did you get the C++ code to train and test your network?

Thanks

[Jayshree Jadhav]

6 Feb

Can you please tell me how to create "mnist_train_index.txt"