Integrate protobuf in matlab.

The separate protobuf (e.g.,
https://gist.github.com/jiayuzhou/b5029bb1ba7bd7f1d911) is likely to
crash Matlab due to the conflict below:

[libprotobuf ERROR google/protobuf/descriptor_database.cc:57] File
already exists in database: caffe.proto
[libprotobuf FATAL google/protobuf/descriptor.cc:1018] CHECK failed:
generated_database_->Add(encoded_file_descriptor, size)

Makefile

@@ -1,5 +1,6 @@
 PROJECT := caffe
 
+
 CONFIG_FILE := Makefile.config
 # Explicitly check for the config file, otherwise make -k will proceed anyway.
 ifeq ($(wildcard $(CONFIG_FILE)),)
@@ -456,7 +457,7 @@ $(MAT$(PROJECT)_SO): $(MAT$(PROJECT)_SRC) $(STATIC_NAME)
 		exit 1; \
 	fi
 	@ echo MEX $<
-	$(Q)$(MATLAB_DIR)/bin/mex $(MAT$(PROJECT)_SRC) \
+	$(Q)$(MATLAB_DIR)/bin/mex -I$(MEXPLUS_DIR) $(MAT$(PROJECT)_SRC) \
 			CXX="$(CXX)" \
 			CXXFLAGS="\$$CXXFLAGS $(MATLAB_CXXFLAGS)" \
 			CXXLIBS="\$$CXXLIBS $(STATIC_LINK_COMMAND) $(LDFLAGS)" -output $@

Makefile.config.example

@@ -45,6 +45,7 @@ BLAS := atlas
 # MATLAB directory should contain the mex binary in /bin.
 # MATLAB_DIR := /usr/local
 # MATLAB_DIR := /Applications/MATLAB_R2012b.app
+# MEXPLUS_DIR := /path/to/your/mexplus/include
 
 # NOTE: this is required only if you will compile the python interface.
 # We need to be able to find Python.h and numpy/arrayobject.h.

matlab/+caffe/fromDatum.m

@@ -0,0 +1,11 @@
+function [ label, image ] = fromDatum( varargin )
+%FROMDATUM decode image and label from caffe protobuf.
+
+CHECK(nargin > 0, ['usage: '...
+    '[ label, image ] = fromDatum( datum )']); 
+datum = varargin{1};
+
+[label, image] = caffe_('from_datum', datum);
+
+end
+

matlab/+caffe/private/caffe_.cpp

@@ -14,12 +14,14 @@
 #include <vector>
 
 #include "mex.h"
-
+#include "mexplus.h"
 #include "caffe/caffe.hpp"
 
 #define MEX_ARGS int nlhs, mxArray **plhs, int nrhs, const mxArray **prhs
 
 using namespace caffe;  // NOLINT(build/namespaces)
+//using namespace std;
+//using namespace mexplus;
 
 // Do CHECK and throw a Mex error if check fails
 inline void mxCHECK(bool expr, const char* msg) {
@@ -478,6 +480,68 @@ static void read_mean(MEX_ARGS) {
   mxFree(mean_proto_file);
 }
 
+// Usage: caffe_('from_datum', datum)
+static void from_datum(MEX_ARGS) {
+  mexplus::OutputArguments output(nlhs, plhs, 2);
+  mxCHECK(nrhs == 1 && mxIsChar(prhs[0]),
+      "Usage: caffe_('from_datum', datum)");
+
+  caffe::Datum datum;
+  std::basic_string<char> datum_content = mexplus::MxArray::to<string>(prhs[0]); //datum received from matlab
+
+  mxCHECK(datum.ParseFromString(datum_content),
+         "Failed to parse datum.");
+  output.set(0, datum.data());
+
+   if (datum.has_encoded() && datum.encoded()) {
+     output.set(0, datum.data());
+   }
+   else {
+    vector<mwIndex> dimensions(3);
+    dimensions[0] = (datum.has_height()) ? datum.height() : 0;
+    dimensions[1] = (datum.has_width()) ? datum.width() : 0;
+    dimensions[2] = (datum.has_channels()) ? datum.channels() : 0;
+    mexplus::MxArray array;
+    vector<mwIndex> subscripts(3);
+    int index = 0;
+    if (datum.has_data()) {
+      array.reset(mxCreateNumericArray(dimensions.size(),
+                                       &dimensions[0],
+                                       mxUINT8_CLASS,
+                                       mxREAL));
+      const string& data = datum.data();
+      for (int k = dimensions[2] - 1; k >= 0; --k) { // BGR to RGB order.
+        subscripts[2] = k;
+        for (int i = 0; i < dimensions[0]; ++i) {
+          subscripts[0] = i;
+          for (int j = 0; j < dimensions[1]; ++j) {
+            subscripts[1] = j;
+            array.set(subscripts, data[index++]);
+          }
+        }
+      }
+    }
+    else if (datum.float_data_size() > 0) {
+      array.reset(mxCreateNumericArray(dimensions.size(),
+                                       &dimensions[0],
+                                       mxSINGLE_CLASS,
+                                       mxREAL));
+      for (int k = dimensions[2] - 1; k >= 0; --k) { // BGR to RGB order.
+        subscripts[2] = k;
+        for (int i = 0; i < dimensions[0]; ++i) {
+          subscripts[0] = i;
+          for (int j = 0; j < dimensions[1]; ++j) {
+            subscripts[1] = j;
+            array.set(subscripts, datum.float_data(index++));
+          }
+        }
+      }
+    }
+    output.set(0, array.release());
+  }
+   output.set(1, (datum.has_label()) ? datum.label() : 0);  
+}
+
 /** -----------------------------------------------------------------
  ** Available commands.
  **/
@@ -515,6 +579,7 @@ static handler_registry handlers[] = {
   { "get_init_key",       get_init_key    },
   { "reset",              reset           },
   { "read_mean",          read_mean       },
+  { "from_datum",         from_datum      },
   // The end.
   { "END",                NULL            },
 };

matlab/demo/lmdb_datum_demo.m

@@ -0,0 +1,72 @@
+% Example of using lmdb and caffe protobuf (datum) in matlab. 
+%
+% by Jiayu, July 1, 2015. 
+%
+% NOTE 1. start matlab with a specified libtiff.5.dylib.
+%    DYLD_INSERT_LIBRARIES=/usr/local/lib/libtiff.5.dylib /Applications/MATLAB_R2012b.app/bin/matlab &
+%
+%      2. install matlab-lmdb 
+%   https://github.com/illidanlab/matlab-lmdb
+%
+%      3. the image num (the first input_num) in the model file should set to 1. 
+%         will fix later. 
+
+if exist('../+caffe', 'dir')
+  addpath('..');
+else
+  error('Please run this demo from caffe/matlab/demo');
+end
+
+addpath ../../../matlab-lmdb/  % change to your matlab-lmdb path
+
+cur_director = pwd;
+net_model    = strcat(cur_director, '/../../examples/mnist/lenet.prototxt');
+net_weights  = strcat(cur_director, '/../../examples/mnist/lenet_iter_10000.caffemodel');
+db_path      = strcat(cur_director, '/../../examples/mnist/mnist_test_lmdb');
+use_gpu = 0;
+phase = 'test';
+
+
+% create caffe net instance
+caffe.set_mode_cpu();
+net = caffe.Net(net_model, net_weights, phase);
+
+% load an existing lmdb database (crated using the shell in example). 
+database = lmdb.DB(db_path, 'RDONLY', true, 'NOLOCK', true);
+cursor = database.cursor('RDONLY', true);
+
+max_count = 10; % maximum test cases
+
+count = 0;      
+correctNum = 0;
+while cursor.next()
+  key = cursor.key;
+  value = cursor.value;
+
+  % transform datum. 
+  [image, label] = caffe.fromDatum(value);
+
+  % prepare image
+  data = single(image);
+  data = permute(data, [2,1,3]);
+
+  % generate prediction 
+  scores = net.forward({data});
+  predict_class = find(scores{1}==1) - 1; % shift 1
+
+
+  fprintf('[%u] Class %u predicted as %u \n', count+1, label, predict_class)
+
+  if(predict_class == label)
+      correctNum = correctNum + 1;
+  end
+
+  count = count + 1;
+  if (count >= max_count)
+      break;
+  end
+end
+
+fprintf('Correctly classified %d images out of %d ( %d percent)\n', correctNum, count, correctNum/count * 100)
+
+clear cursor;