如何在c ++中加载caffe模型进行预测

Question

I have been using Caffe on Python so far and now I am trying to use C++ to familiarize myself.

What I have done is I tried to explore the caffe FC layers by computing features and loading thru HDF5 layer. I have trained the model and it works very well with python using the following code:

caffe.set_device(0)
caffe.set_mode_gpu()
net = caffe.Net(proto_file, caffe_model, caffe.TEST)    
feats, labels = get_features('test/test.txt')   #AlexNet features
for feature, label in zip(feats, labels):
    net.blobs['data'].data[...] = feature
    output = net.forward()
    output_prob = output['loss'][0]
    print output_prob.argmax(), ", ", label

Using this python code I can check and it works very well.

I am trying to write the code in c++ to do the same prediction. This line

net.blobs['data'].data[...] = feature

is bit tricky and I cannot do the same in c++: How can I load features into the data layer in c++:

my c++ code so far is:

    caffe::Caffe::SetDevice(0);
    caffe::Caffe::set_mode(caffe::Caffe::GPU);
    boost::shared_ptr<caffe::Net<float> > net_;
    net_.reset(new caffe::Net<float>(model_file, caffe::TEST));
    net_->CopyTrainedLayersFrom(trained_file);

    std::cout << "LOADED CAFFE MODEL\n";
    LOG(INFO) << "Blob size: "<< net_->input_blobs().size();

This caffe example is useful but it loads image and then separates the channels. In my case I have 4096-D feature vector from AlexNet which I want to load directly as in Python code.

Answer 1

Get the blob index according to its name:

const std::vector<std::string>& blob_names_ = net_->blob_names();
auto it = std::find(blob_names_.begin(), blob_names_.end(), "data");
int index = -1;
if (it == blob_names_.end())
{
    // no "data" blob, do error handling
}
else
{
    index = std::distance(blob_names_.begin(), it);
}

(Of course you can also get the index by iterating over blob_names_ and compare each item).

Change the blob data:

float* feature = your_func_to_get_feature(); // custom this
caffe::Blob<float>*>& blob_to_set_ = net_->blobs()[index];
blob_to_set_->set_cpu_data(feature);

Continue from this point as usual.