can't feed the data to the caffe code that i have !

[Mehrdad HMT]

Hi everyone.

I'm working on some project about image tagging and I'm new to caffe. I have a caffe code from one of the papers that i read about image tagging and the data set that is used for it, is CIFAR100.

According to code and layers structure, I have to feed image features in LMDB format and labels in HDF5 format into the code separately. I converted the features into the lmdb format and labels into the hdf5 format but yet i can't run the code. I think it should be some preprocessing on the labels before feed them into the network but i couldn't understand how! I wrote an email to paper's Authors but unfortunately they didn't respond yet.

as the paper author told me, the data set should be in a hierarchical structure so I'm trying to use CIFAR100 data set with two-level hierarchical structure: every image has to labels, a coarse label and a fine label.

I put the code here, so if anybody could run it with the CIFAR100 dataset, please teach me how.

this is proto.txt file:

name: "res_to_lstm"

layer {
  name: "image_feature"
  type: "Data"
  top: "global_pool"
  include { phase: TRAIN }
  data_param {
    source: "./examples/coarse-to-fine/Imagenet/ResNet/feature/global_pool_train"
    batch_size: 100
    backend: LMDB
  }
}

layer {
  name: "data"
  type: "HDF5Data"
  top: "cont_sentence"
  top: "input_sentence"
  top: "target_sentence"
  include { phase: TRAIN }
  hdf5_data_param {
    source: "/home/destiny/Datasets/Imagenet/train_h5_caffe/hdf5_chunk_list_shuffle.txt"
    batch_size: 10
  }
}

layer {
  name: "image_feature"
  type: "Data"
  top: "global_pool"
  include { phase: TEST }
  data_param {
    source: "./examples/coarse-to-fine/Imagenet/ResNet/feature/global_pool_val"
    batch_size: 100
    backend: LMDB
  }
}

layer {
  name: "data"
  type: "HDF5Data"
  top: "cont_sentence"
  top: "input_sentence"
  top: "target_sentence"
  include { phase: TEST }
  hdf5_data_param {
    source: "/home/destiny/Datasets/Imagenet/val_h5_caffe/hdf5_chunk_list.txt"
    batch_size: 10
  }
}

layer {
  name: "embedding"
  type: "Embed"
  bottom: "input_sentence"
  top: "embedded_input_sentence"
  param {
    lr_mult: 1
  }
  embed_param {
    bias_term: false
    input_dim: 1861
    num_output: 1000
    weight_filler {
      type: "uniform"
      min: -0.08
      max: 0.08
    }
  }
}
layer {
  name: "lstm1"
  type: "LSTM"
  bottom: "embedded_input_sentence"
  bottom: "cont_sentence"
  bottom: "global_pool"
  top: "lstm1"
  recurrent_param {
    num_output: 1000
    weight_filler {
      type: "uniform"
      min: -0.08
      max: 0.08
    }
    bias_filler {
      type: "constant"
      value: 0
    }
  }
}

layer {
  name: "predict"
  type: "InnerProduct"
  bottom: "lstm1"
  top: "predict"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  inner_product_param {
    num_output: 1861
    weight_filler {
      type: "uniform"
      min: -0.08
      max: 0.08
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    axis: 2
  }
}

layer {
  name: "cross_entropy_loss"
  type: "SoftmaxWithLoss"
  bottom: "predict"
  bottom: "target_sentence"
  top: "cross_entropy_loss"
  loss_weight: 10
  loss_param {
    ignore_label: -1
  }
  softmax_param {
    axis: 2
  }
}
layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "predict"
  bottom: "target_sentence"
  top: "accuracy"
  include { phase: TEST }
  accuracy_param {
    axis: 2
    ignore_label: -1
  }
}

--------------------------------------------------------

and this is the solver.prototxt file:

net: "./examples/coarse-to-fine/Imagenet/ResNet-BN/ResNet_train_lstm.prototxt"
test_iter: 500
test_interval: 1000
test_initialization: true
base_lr: 0.1
lr_policy: "step"
gamma: 0.5
stepsize: 60000
display: 200
max_iter: 260000
momentum: 0.9
weight_decay: 0.0000
snapshot: 10000
snapshot_prefix: "./examples/coarse-to-fine/Imagenet/ResNet/models/global_pool_lstm"
solver_mode: GPU
random_seed: 1701
average_loss: 100
clip_gradients: 10

---

CNN-RNN: A Unified Framework for Multi-label Image Classification

[ALI RAZA]

sorry to ask it might looks irrelevant, but i want to ask you are you using caffe olny are segnet-caffe? i am new BS student if you guide me i will be very very thank you. plz help me