Add a new option to add a labels file and a means file.

src/Simulator.cpp

@@ -1,5 +1,5 @@
+#include <cassert>
 #include <iostream>
-#include <iterator>
 #include <vector>
 
 #include "Simulator.hpp"
@@ -8,7 +8,7 @@ using namespace caffe;
 using namespace std;
 using namespace fmri;
 
-Simulator::Simulator(const string& model_file, const string& weights_file) :
+Simulator::Simulator(const string& model_file, const string& weights_file, const string& means_file) :
 	net(model_file, TEST)
 {
 	net.CopyTrainedLayersFrom(weights_file);
@@ -21,16 +21,18 @@ Simulator::Simulator(const string& model_file, const string& weights_file) :
 			input_geometry.height, input_geometry.width);
 	/* Forward dimension change to all layers. */
 	net.Reshape();
+
+    if (means_file != "")  {
+        means = processMeans(means_file);
+    }
+
 }
 
-void Simulator::simulate(const string& image_file)
+vector<Simulator::DType> Simulator::simulate(const string& image_file)
 {
 	cv::Mat im = cv::imread(image_file, -1);
 
-	if (im.empty()) {
-		cerr << "Unable to read " << image_file << endl;
-		return;
-	}
+    assert(!im.empty());
 
     auto input = preprocess(im);
     auto channels = getWrappedInputLayer();
@@ -44,10 +46,7 @@ void Simulator::simulate(const string& image_file)
     const DType *end = begin + output_layer->channels();
     vector<DType> result(begin, end);
 
-    // TODO: visualize, rather than just print.
-    for (auto v : result) {
-        cout << v << endl;
-    }
+    return result;
 }
 
 vector<cv::Mat> Simulator::getWrappedInputLayer()
@@ -111,8 +110,36 @@ cv::Mat Simulator::preprocess(cv::Mat original) const
     cv::Mat sample_float;
     resized.convertTo(sample_float, num_channels == 3 ? CV_32FC3 : CV_32FC1);
 
-    // TODO: substract means.
-    // Don't know if necessary yet.
+    if (means.empty()) {
+        return sample_float;
+    }
+
+    cv::Mat normalized;
+    cv::subtract(sample_float, means, normalized);
+
+    return normalized;
 
-    return sample_float;
+}
+
+cv::Mat Simulator::processMeans(const string &means_file) const
+{
+    BlobProto proto;
+    ReadProtoFromBinaryFileOrDie(means_file, &proto);
+
+    Blob<DType> mean_blob;
+    mean_blob.FromProto(proto);
+
+    assert(mean_blob.channels() == num_channels);
+
+    vector<cv::Mat> channels;
+    float* data = mean_blob.mutable_cpu_data();
+    for (unsigned int i = 0; i < num_channels; ++i) {
+        channels.emplace_back(mean_blob.height(), mean_blob.width(), CV_32FC1, data);
+        data += mean_blob.height() * mean_blob.width();
+    }
+
+    cv::Mat mean;
+    cv::merge(channels, mean);
+
+    return cv::Mat(input_geometry, mean.type(), cv::mean(mean));
 }