diff --git a/CMakeLists.txt b/CMakeLists.txt
index 760774b6b..62edeb331 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -41,6 +41,7 @@ caffe_option(USE_LEVELDB "Build with levelDB" ON)
 caffe_option(USE_LMDB "Build with lmdb" ON)
 caffe_option(ALLOW_LMDB_NOLOCK "Allow MDB_NOLOCK when reading LMDB files (only if necessary)" OFF)
 caffe_option(USE_OPENMP "Link with OpenMP (when your BLAS wants OpenMP and you get linker errors)" OFF)
+caffe_option(HAVE_BINGING "caffe binding" ON)
 
 # ---[ Dependencies
 include(cmake/Dependencies.cmake)
@@ -50,6 +51,8 @@ if(UNIX OR APPLE)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fPIC -Wall")
 endif()
 
+SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -D_MWAITXINTRIN_H_INCLUDED")
+
 caffe_set_caffe_link()
 
 if(USE_libstdcpp)
@@ -107,6 +110,10 @@ add_subdirectory(python)
 add_subdirectory(matlab)
 add_subdirectory(docs)
 
+if (HAVE_BINGING)
+  add_subdirectory(windows/caffe.binding)
+endif()
+
 # ---[ Linter target
 add_custom_target(lint COMMAND ${CMAKE_COMMAND} -P ${PROJECT_SOURCE_DIR}/cmake/lint.cmake)
 
diff --git a/README.md b/README.md
index 7a7cbff50..22981f416 100644
--- a/README.md
+++ b/README.md
@@ -64,6 +64,12 @@ After you have built solution with Matlab support, in order to use it you have t
 ### Build
 Now, you should be able to build `.\windows\Caffe.sln`
 
+
+## Ubuntu 16.04 cmake build 
+`cmake . -DCUDA_NVCC_FLAGS="-D_FORCE_INLINES"`
+
+`make`
+
 ## License and Citation
 
 Caffe is released under the [BSD 2-Clause license](https://github.com/BVLC/caffe/blob/master/LICENSE).
diff --git a/python/caffe/_caffe.cpp b/python/caffe/_caffe.cpp
index 8473d64e1..a1479f387 100644
--- a/python/caffe/_caffe.cpp
+++ b/python/caffe/_caffe.cpp
@@ -188,7 +188,7 @@ namespace caffe {
                           bp::object labels_obj) {
     // check that this network has an input MemoryDataLayer
     shared_ptr<MemoryDataLayer<Dtype> > md_layer =
-      boost::dynamic_pointer_cast<MemoryDataLayer<Dtype>>(net->layers()[0]);
+      boost::dynamic_pointer_cast<MemoryDataLayer<Dtype> >(net->layers()[0]);
     if (!md_layer) {
       throw std::runtime_error("set_input_arrays may only be called if the"
                                " first layer is a MemoryDataLayer");
@@ -588,4 +588,4 @@ namespace caffe {
     import_array1();
   }
 
-}  // namespace caffe
\ No newline at end of file
+}  // namespace caffe
diff --git a/src/caffe/common.cpp b/src/caffe/common.cpp
index 06d3637ac..36109657b 100644
--- a/src/caffe/common.cpp
+++ b/src/caffe/common.cpp
@@ -139,10 +139,11 @@ void* Caffe::RNG::generator() {
 #else  // Normal GPU + CPU Caffe.
 
 Caffe::Caffe()
-    : cublas_handle_(NULL), curand_generator_(NULL), random_generator_(),
+    : cublas_handle_(NULL), curand_generator_(NULL), 
 #ifdef USE_CUDNN
-  cudnn_handle_(NULL),
+    cudnn_handle_(NULL),
 #endif
+    random_generator_(),
     mode_(Caffe::CPU),
     solver_count_(1), solver_rank_(0), multiprocess_(false) {
   // Try to create a cublas handler, and report an error if failed (but we will
diff --git a/src/caffe/layers/batch_contrastive_loss_layer.cpp b/src/caffe/layers/batch_contrastive_loss_layer.cpp
index 912abe5af..0ce70d243 100644
--- a/src/caffe/layers/batch_contrastive_loss_layer.cpp
+++ b/src/caffe/layers/batch_contrastive_loss_layer.cpp
@@ -26,7 +26,13 @@ void BatchContrastiveLossLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& botto
   CHECK_EQ(bottom[0]->num(), bottom[0]->channels());
   if (top.size() >= 2) {
     // positive distance, negative distance.
+#if __cplusplus < 201103L
+    int arr[] = { 2 };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[1]->Reshape(shape);
+#else
     top[1]->Reshape({ 2 });
+#endif
   }
 }
 
@@ -34,7 +40,7 @@ template <typename Dtype>
 void BatchContrastiveLossLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
     const vector<Blob<Dtype>*>& top) {
   const Dtype* bottom_data = bottom[0]->cpu_data();
-  Dtype* bottom_diff = bottom[0]->mutable_cpu_diff();
+//  Dtype* bottom_diff = bottom[0]->mutable_cpu_diff();
   const Dtype* label = bottom[1]->cpu_data();
   int num = bottom[0]->num();
   Dtype positive_distance = Dtype(0);
diff --git a/src/caffe/layers/general_constrastive_loss_layer.cpp b/src/caffe/layers/general_constrastive_loss_layer.cpp
index f84250714..8bcfbdefd 100644
--- a/src/caffe/layers/general_constrastive_loss_layer.cpp
+++ b/src/caffe/layers/general_constrastive_loss_layer.cpp
@@ -36,15 +36,33 @@ void GeneralContrastiveLossLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bot
   if (top.size() >= 2) {
     if (add_intra_mae_) {
       // positive distance, negative distance, intra_mae.
+#if __cplusplus < 201103L
+      int arr[] = { 3 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ 3 });
+#endif
     }
     else {
       // positive distance, negative distance.
+#if __cplusplus < 201103L
+      int arr[] = { 2 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ 2 });
+#endif
     }
   }
   if (max_negative_only_) {
+#if __cplusplus < 201103L
+    int arr[] = { bottom[0]->num() };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    max_negative_index_.Reshape(shape);
+#else
     max_negative_index_.Reshape({ bottom[0]->num() });
+#endif
   }
 }
 
@@ -59,7 +77,7 @@ void GeneralContrastiveLossLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>&
   int num = bottom[0]->num();
   int count = bottom[0]->count();
   int dim = count / num;
-  Dtype weighted_count = num * (abs(positive_weight_) + (dim - 1)*abs(negative_weight_));
+//  Dtype weighted_count = num * (abs(positive_weight_) + (dim - 1)*abs(negative_weight_));
   Dtype positive_distance = Dtype(0);
   Dtype negative_distance = Dtype(0);
   max_positive_index_ = 0;
diff --git a/src/caffe/layers/general_triplet_loss_layer.cpp b/src/caffe/layers/general_triplet_loss_layer.cpp
index 46333d541..5a17afbc8 100644
--- a/src/caffe/layers/general_triplet_loss_layer.cpp
+++ b/src/caffe/layers/general_triplet_loss_layer.cpp
@@ -27,10 +27,22 @@ void GeneralTripletLossLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
   LossLayer<Dtype>::Reshape(bottom, top);
   if (top.size() >= 2) {
     // positive distance, negative distance.
+#if __cplusplus < 201103L
+    int arr[] = { 2 };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[1]->Reshape(shape);
+#else
     top[1]->Reshape({ 2 });
+#endif
   }
   if (hardest_only_) {
+#if __cplusplus < 201103L
+    int arr[] = { bottom[0]->num() };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    hardest_index_.Reshape(shape);
+#else
     hardest_index_.Reshape({ bottom[0]->num() });
+#endif
   }
 }
 
diff --git a/src/caffe/layers/hotspot_layer.cpp b/src/caffe/layers/hotspot_layer.cpp
index 6c0ab7363..c5e7641e5 100644
--- a/src/caffe/layers/hotspot_layer.cpp
+++ b/src/caffe/layers/hotspot_layer.cpp
@@ -38,7 +38,13 @@ void HotspotLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
     height_ = bottom[1]->height();
     width_ = bottom[1]->width();
   }
+#if __cplusplus < 201103L
+    int arr[] = { bottom[0]->num(), num_point, height_, width_ };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[0]->Reshape(shape);
+#else
   top[0]->Reshape({ bottom[0]->num(), num_point, height_, width_ });
+#endif
 }
 
 template <typename Dtype>
diff --git a/src/caffe/layers/image_data_layer.cpp b/src/caffe/layers/image_data_layer.cpp
index 00df2bb43..92e5f4741 100644
--- a/src/caffe/layers/image_data_layer.cpp
+++ b/src/caffe/layers/image_data_layer.cpp
@@ -55,9 +55,16 @@ void ImageDataLayer<Dtype>::DataLayerSetUp(const vector<Blob<Dtype>*>& bottom,
   if (top.size() == 3) {
     num_samples_ = vector<int>(max_label + 1);
     class_weights_ = vector<Dtype>(max_label + 1);
+#if __cplusplus < 201103L
+    for (vector<std::pair<std::string, int> >::iterator l  = lines_.begin(); l != lines_.end(); ++l) {
+      num_samples_[l->second]++;
+    }
+#else
     for (auto l : lines_) {
       num_samples_[l.second]++;
     }
+#endif
+
     Dtype mean_sample_num = (Dtype)lines_.size() / (Dtype)(max_label + 1);
     Dtype min_weight = 9999, max_weight = 0;
     for (int i = 0; i < num_samples_.size(); i++) {
@@ -76,10 +83,17 @@ void ImageDataLayer<Dtype>::DataLayerSetUp(const vector<Blob<Dtype>*>& bottom,
   if (balance_) {
     num_samples_ = vector<int>(max_label + 1);
     filename_by_class_ = vector<vector<std::pair<std::string, int> > >(max_label + 1);
+#if __cplusplus < 201103L
+    for (vector<std::pair<std::string, int> >::iterator l  = lines_.begin(); l != lines_.end(); ++l) {
+      num_samples_[l->second]++;
+      filename_by_class_[l->second].push_back(std::make_pair(l->first, 0));
+    }
+#else
     for (auto l : lines_) {
       num_samples_[l.second]++;
       filename_by_class_[l.second].push_back(std::make_pair(l.first, 0));
     }
+#endif
     class_id_ = 0;
   }
 
@@ -194,6 +208,20 @@ void ImageDataLayer<Dtype>::load_batch(Batch<Dtype>* batch) {
 
       if (balance_) {
         int pick_index = (caffe_rng_rand() % num_samples_[class_id_]) + 1;
+#if __cplusplus < 201103L
+        vector<std::pair<std::string, int> >* samples = &filename_by_class_[class_id_];
+        for (vector<std::pair<std::string, int> >::iterator sample = samples->begin(); sample != samples->end(); ++sample) {
+          if (sample->second == 0) {
+            pick_index--;
+            if (pick_index == 0) {
+              this_line = std::make_pair(sample->first, class_id_);
+              sample->second = 1;
+              num_samples_[class_id_]--;
+              break;
+            }
+          }
+        }
+#else
         for (auto& sample : filename_by_class_[class_id_]) {
           if (sample.second == 0) {
             pick_index--;
@@ -205,12 +233,20 @@ void ImageDataLayer<Dtype>::load_batch(Batch<Dtype>* batch) {
             }
           }
         }
+#endif
         CHECK_GT(this_line.first.size(), 0);
         if (num_samples_[class_id_] == 0) {
           num_samples_[class_id_] = filename_by_class_[class_id_].size();
+#if __cplusplus < 201103L
+          vector<std::pair<std::string, int> >* samples = &filename_by_class_[class_id_];
+          for (vector<std::pair<std::string, int> >::iterator sample = samples->begin(); sample != samples->end(); ++sample) {
+            sample->second = 0;
+          }
+#else
           for (auto& sample : filename_by_class_[class_id_]) {
             sample.second = 0;
           }
+#endif
         }
       }
       else {
diff --git a/src/caffe/layers/insanity_layer.cpp b/src/caffe/layers/insanity_layer.cpp
index f0d25b285..93615891a 100644
--- a/src/caffe/layers/insanity_layer.cpp
+++ b/src/caffe/layers/insanity_layer.cpp
@@ -13,7 +13,7 @@ void InsanityLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
   CHECK_GE(bottom[0]->num_axes(), 2)
       << "Number of axes of bottom blob must be >=2.";
   InsanityParameter insanity_param_ = this->layer_param().insanity_param();
-  int channels = bottom[0]->channels();
+ // int channels = bottom[0]->channels();
   lb_ = insanity_param_.lb();
   ub_ = insanity_param_.ub();
   CHECK_GT(ub_, lb_) << "upper bound must > lower bound.";
@@ -40,8 +40,8 @@ void InsanityLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
   const Dtype* bottom_data = bottom[0]->cpu_data();
   Dtype* top_data = top[0]->mutable_cpu_data();
   const int count = bottom[0]->count();
-  const int dim = bottom[0]->count(2);
-  const int channels = bottom[0]->channels();
+//  const int dim = bottom[0]->count(2);
+//  const int channels = bottom[0]->channels();
   Dtype* slope_data = alpha.mutable_cpu_data();
 
   // For in-place computation
@@ -73,8 +73,8 @@ void InsanityLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
   const Dtype* slope_data = alpha.cpu_data();
   const Dtype* top_diff = top[0]->cpu_diff();
   const int count = bottom[0]->count();
-  const int dim = bottom[0]->count(2);
-  const int channels = bottom[0]->channels();
+//  const int dim = bottom[0]->count(2);
+//  const int channels = bottom[0]->channels();
 
   // For in-place computation
   if (top[0] == bottom[0] && lb_ < 0) {
diff --git a/src/caffe/layers/label_specific_rescale.cpp b/src/caffe/layers/label_specific_rescale.cpp
index 2ebe87030..32de73ee4 100644
--- a/src/caffe/layers/label_specific_rescale.cpp
+++ b/src/caffe/layers/label_specific_rescale.cpp
@@ -55,7 +55,7 @@ void LabelSpecificRescaleLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>&
                                                     const vector<bool>& propagate_down,
                                                     const vector<Blob<Dtype>*>& bottom) {
   if (propagate_down[0]) {
-    const Dtype* bottom_data = bottom[0]->cpu_data();
+//    const Dtype* bottom_data = bottom[0]->cpu_data();
     const Dtype* label_data = bottom[1]->cpu_data();
     const Dtype* top_diff = top[0]->cpu_diff();
     Dtype* bottom_diff = bottom[0]->mutable_cpu_diff();
diff --git a/src/caffe/layers/multi_label_image_data_layer.cpp b/src/caffe/layers/multi_label_image_data_layer.cpp
index a419c8693..e404b9768 100644
--- a/src/caffe/layers/multi_label_image_data_layer.cpp
+++ b/src/caffe/layers/multi_label_image_data_layer.cpp
@@ -8,7 +8,11 @@
 #include <string>
 #include <utility>
 #include <vector>
+#if __cplusplus >= 201103L
 #include <random>
+#else
+#include <boost/random.hpp>
+#endif
 
 #include "caffe/layers/multi_label_image_data_layer.hpp"
 #include "caffe/util/benchmark.hpp"
@@ -23,7 +27,11 @@ namespace caffe {
     this->StopInternalThread();
   }
 
+#if __cplusplus >= 201103L
   typedef std::mt19937 RANDOM_ENGINE;
+#else
+  typedef boost::mt19937 RANDOM_ENGINE;
+#endif
 
   template <typename Dtype>
   void extract_face(cv::Mat& input_image, Dtype* points, int point_count,
@@ -40,14 +48,27 @@ namespace caffe {
     double face_scale = 2 * sqrt((face_center.x - mouth_center.x) * (face_center.x - mouth_center.x)
                                  + (face_center.y - mouth_center.y) * (face_center.y - mouth_center.y));
     RANDOM_ENGINE prnd(time(NULL));
+
+#if __cplusplus >= 201103L
     face_center.x += std::uniform_int_distribution<int>(-max_random_shift, max_random_shift)(prnd);
     face_center.y += std::uniform_int_distribution<int>(-max_random_shift, max_random_shift)(prnd);
     std::uniform_real_distribution<float> rand_uniform(0, 1);
+#else
+    face_center.x += boost::random::uniform_int_distribution<int>(-max_random_shift, max_random_shift)(prnd);
+    face_center.y += boost::random::uniform_int_distribution<int>(-max_random_shift, max_random_shift)(prnd);
+    boost::random::uniform_real_distribution<float> rand_uniform(0, 1);
+#endif
+
     // shear
     float s = rand_uniform(prnd) * max_shear_ratio * 2 - max_shear_ratio;
     // rotate
+#if __cplusplus >= 201103L
     int angle = std::uniform_int_distribution<int>(
       -max_rotate_angle, max_rotate_angle)(prnd);
+#else
+    int angle = boost::random::uniform_int_distribution<int>(
+      -max_rotate_angle, max_rotate_angle)(prnd);
+#endif
     float a = cos(angle / 180.0 * CV_PI);
     float b = sin(angle / 180.0 * CV_PI);
     // scale
@@ -61,7 +82,11 @@ namespace caffe {
     float ws = ratio * hs;
     int flip = 1;
     if (face_mirror) {
+#if __cplusplus >= 201103L
       flip = std::uniform_int_distribution<int>(0, 1)(prnd)* 2 - 1;
+#else
+      flip = boost::random::uniform_int_distribution<int>(0, 1)(prnd)* 2 - 1;
+#endif
     }
     hs *= flip;
 
@@ -149,10 +174,17 @@ namespace caffe {
     if (balance_) {
       num_samples_ = vector<int>(max_label + 1);
       filename_by_class_ = vector<vector<std::pair<std::string, shared_ptr<vector<Dtype> > > > >(max_label + 1);
+#if __cplusplus >= 201103L
       for (auto& l : lines_) {
         num_samples_[(*l.second)[balance_by_]]++;
         filename_by_class_[(*l.second)[balance_by_]].push_back(l);
       }
+#else
+      for (typename vector<std::pair<std::string, shared_ptr<vector<Dtype> > > >::iterator l  = lines_.begin(); l != lines_.end(); ++l) {
+        num_samples_[(*l->second)[balance_by_]]++;
+        filename_by_class_[(*l->second)[balance_by_]].push_back(*l);
+      }
+#endif
       class_id_ = 0;
     }
 
@@ -194,7 +226,12 @@ namespace caffe {
       << top[0]->channels() << "," << top[0]->height() << ","
       << top[0]->width();
     // label
+#if __cplusplus >= 201103L
     vector<int> label_shape = { batch_size - label_cut_start_ - label_cut_end_, label_count };
+#else
+    int arr[] = { batch_size - label_cut_start_ - label_cut_end_, label_count };
+    vector<int> label_shape(arr,arr+sizeof(arr)/sizeof(int));
+#endif
     top[1]->Reshape(label_shape);
     for (int i = 0; i < this->prefetch_.size(); ++i) {
       this->prefetch_[i]->label_.Reshape(label_shape);
diff --git a/src/caffe/layers/nca_loss_layer.cpp b/src/caffe/layers/nca_loss_layer.cpp
index 7d0226092..6c98dafb3 100644
--- a/src/caffe/layers/nca_loss_layer.cpp
+++ b/src/caffe/layers/nca_loss_layer.cpp
@@ -19,10 +19,22 @@ void NCALossLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
   LossLayer<Dtype>::Reshape(bottom, top);
   if (top.size() == 2) {
     // positive distance, negative distance.
+#if __cplusplus < 201103L
+    int arr[] = { 2 };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[1]->Reshape(shape);
+#else
     top[1]->Reshape({ 2 });
+#endif
   }
   if (min_negative_only_) {
+#if __cplusplus < 201103L
+    int arr[] = { bottom[0]->num() };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    min_negative_index_.Reshape(shape);
+#else
     min_negative_index_.Reshape({ bottom[0]->num() });
+#endif
   }
 }
 
@@ -85,7 +97,7 @@ void NCALossLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
     int count = bottom[0]->count();
     int dim = count / num;
 
-    Dtype negative_sum = Dtype(0);
+//    Dtype negative_sum = Dtype(0);
 
     for (int i = 0; i < num; ++i) {
       if (min_negative_only_) {
diff --git a/src/caffe/layers/normalize_layer.cpp b/src/caffe/layers/normalize_layer.cpp
index 8a1c9f341..ffbf1a401 100644
--- a/src/caffe/layers/normalize_layer.cpp
+++ b/src/caffe/layers/normalize_layer.cpp
@@ -131,4 +131,4 @@ STUB_GPU(NormalizeLayer);
 INSTANTIATE_CLASS(NormalizeLayer);
 REGISTER_LAYER_CLASS(Normalize);
 
-}  // namespace caffe
\ No newline at end of file
+}  // namespace caffe
diff --git a/src/caffe/layers/pairwise_layer.cpp b/src/caffe/layers/pairwise_layer.cpp
index 55ba5e6f0..161e70881 100644
--- a/src/caffe/layers/pairwise_layer.cpp
+++ b/src/caffe/layers/pairwise_layer.cpp
@@ -33,7 +33,13 @@ void PairwiseLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
   M_ = bottom[0]->num();
   N_ = bottom[1]->num();
   K_ = bottom[0]->channels();
+#if __cplusplus < 201103L
+  int arr[] = { M_, N_, K_ };
+  vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+  top[0]->Reshape(shape);
+#else
   top[0]->Reshape({ M_, N_, K_ });
+#endif
   // If max operation, we will initialize the vector index part.
   if (this->layer_param_.pairwise_param().operation() ==
       PairwiseParameter_PairwiseOp_MAX && top.size() == 1) {
@@ -96,8 +102,8 @@ template <typename Dtype>
 void PairwiseLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
     const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
   const int* mask = NULL;
-  const int count = top[0]->count();
-  const Dtype* top_data = top[0]->cpu_data();
+ // const int count = top[0]->count();
+//  const Dtype* top_data = top[0]->cpu_data();
   const Dtype* top_diff = top[0]->cpu_diff();
   const Dtype* bottom_data_a = bottom[0]->cpu_data();
   const Dtype* bottom_data_b = bottom[1]->cpu_data();
diff --git a/src/caffe/layers/pooling_layer.cu b/src/caffe/layers/pooling_layer.cu
index 59d624f07..ae23915b4 100644
--- a/src/caffe/layers/pooling_layer.cu
+++ b/src/caffe/layers/pooling_layer.cu
@@ -378,10 +378,10 @@ if ((n==0)&&( top_data[n*channels_+ch] > 10000))
 			  int phS = ph * stride_;
 		//	 for (int pw = 0; pw < pooled_width_; ++pw) {
 				  int pwS = pw * stride_;
-				  int defval = phS * width_ + pwS;
+//				  int defval = phS * width_ + pwS;
 				 // LOG(INFO) << ph <<","<< pw;
-				  int ihval = Ih[(n*channels_+ch)*N_+phS * width_ + pwS];
-				  int ivval = Iv[(n*channels_+ch)*N_+phS * width_ + pwS];
+//				  int ihval = Ih[(n*channels_+ch)*N_+phS * width_ + pwS];
+//				  int ivval = Iv[(n*channels_+ch)*N_+phS * width_ + pwS];
 				  Dtype hdif = phS - Ih[(n*channels_+ch)*N_+phS * width_ + pwS];
 				  Dtype vdif = pwS - Iv[(n*channels_+ch)*N_+phS * width_ + pwS];
 				//  LOG(INFO) << "fp data[" << defval << "]:" << data_pointer[defval] <<","<< Mdt[defval];
@@ -495,7 +495,7 @@ LOG(INFO) << "Forward_gpu PoolingParameter_PoolMethod_DEF 2.4";
 			  int phS = ph * stride_;
 			  for (int pw = 0; pw < pooled_width_; ++pw) {
 				  int pwS = pw * stride_;
-				  int defval = phS * width_ + pwS;
+//				  int defval = phS * width_ + pwS;
 				  int ihval = Ih[(n*channels_+ch)*N_+phS * width_ + pwS];
 				  int ivval = Iv[(n*channels_+ch)*N_+phS * width_ + pwS];
 				  Dtype hdif = phS - Ih[(n*channels_+ch)*N_+phS * width_ + pwS];
@@ -568,8 +568,8 @@ LOG(INFO) << "Forward_gpu PoolingParameter_PoolMethod_DEF 2.4";
     for (int n = 0; n < bottom[0]->num(); ++n) {
        
 	  int N2 = pooled_height_*pooled_width_;
-	  int center = kernel_size_/2;
-	  int def_center = center+center*kernel_size_;
+//	  int center = kernel_size_/2;
+//	  int def_center = center+center*kernel_size_;
       
 
       for (int ch = 0; ch < channels_; ++ch) {          
@@ -617,9 +617,9 @@ LOG(INFO) << "Forward_gpu PoolingParameter_PoolMethod_DEF 2.4";
 			  if (phS > height_-kernel_size_)
 				  phS = height_ - kernel_size_;
 			  for (int pw = 0; pw < pooled_width_; ++pw) {
-                  int pwS = pw * stride_;
-				  int defval = phS * width_ + pwS;
-				  int defwidx;
+//				  int pwS = pw * stride_;
+//				  int defval = phS * width_ + pwS;
+//				  int defwidx;
 				  int maxIdx;
                   
                   int hstart = ph * stride_h_ - pad_h_;
@@ -944,7 +944,7 @@ void PoolingLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top,
   int* defp;
   int N2;
   int Nparam = kernel_size_*kernel_size_;
-  const Dtype* bottom_data_p;
+//  const Dtype* bottom_data_p;
 
   const Dtype* top_mask = NULL;
   switch (this->layer_param_.pooling_param().pool()) {
@@ -1196,25 +1196,25 @@ void PoolingLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top,
   case PoolingParameter_PoolMethod_DEF_ALL4:
 //LOG(INFO) << "Backward_gpu PoolingParameter_PoolMethod_DEF_ALL3";
 	bottom_diff = bottom[0]->mutable_cpu_diff();
-    top_diff = top[0]->cpu_diff();
-    Ih = Ih_.data();
-    Iv = Iv_.data();
-    defp = defp_.data();
-    d_defw = this->blobs_[0]->mutable_cpu_diff();
-    memset(d_defw, 0, channels_*kernel_size_*kernel_size_* sizeof(Dtype));
-    dh = dh_.mutable_cpu_data();
-    dv = dv_.mutable_cpu_data();
+	top_diff = top[0]->cpu_diff();
+	Ih = Ih_.data();
+	Iv = Iv_.data();
+	defp = defp_.data();
+	d_defw = this->blobs_[0]->mutable_cpu_diff();
+	memset(d_defw, 0, channels_*kernel_size_*kernel_size_* sizeof(Dtype));
+	dh = dh_.mutable_cpu_data();
+	dv = dv_.mutable_cpu_data();
 	N2 = pooled_width_*pooled_height_;
-    for (int n = 0; n < top[0]->num(); ++n) {
-      for (int ch = 0; ch < channels_; ++ch) {
+	for (int n = 0; n < top[0]->num(); ++n) {
+              for (int ch = 0; ch < channels_; ++ch) {
 /*            int vstart = Iv[(n*channels_+ch)*N_+defp[ch]];
             int hstart = Ih[(n*channels_+ch)*N_+defp[ch]];
             int vend = min(vstart + kernel_size_, width_);
             int hend = min(hstart + kernel_size_, height_);*/
 		  for (int ph = 0; ph < pooled_height_; ++ph) {
-			  int phS = ph * stride_;
+			  //int phS = ph * stride_;
 			  for (int pw = 0; pw < pooled_width_; ++pw) {
-				  int pwS = pw * stride_;
+//				  int pwS = pw * stride_;
 				  int idxInv = Ih[(n*channels_+ch)*N2+ph * pooled_width_ + pw];
 				  Dtype t_dif = top_diff[ph * pooled_width_ + pw];
 				  bottom_diff[idxInv] += t_dif;
diff --git a/src/caffe/layers/predict_box_layer.cpp b/src/caffe/layers/predict_box_layer.cpp
index d8150f11f..33daffdda 100644
--- a/src/caffe/layers/predict_box_layer.cpp
+++ b/src/caffe/layers/predict_box_layer.cpp
@@ -47,12 +47,25 @@ void PredictBoxLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
     CHECK_EQ(bottom[0]->height(), bottom[2]->height());
     CHECK_EQ(bottom[0]->width(), bottom[2]->width());
   }
-
+#if __cplusplus < 201103L
+  int arr[] = { bottom[0]->num(), 5, bottom[0]->height(), bottom[0]->width() };
+  vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+  top[0]->Reshape(shape);
+  if (output_vector_) {
+    int arr_output[] = { bottom[0]->num(), 1, 5 };
+    vector<int> shape_output(arr_output,arr_output+sizeof(arr_output)/sizeof(int));
+    top[1]->Reshape(shape_output);//will be modified on the fly.
+  }
+  int arr_counter[] = { bottom[0]->num(),1,bottom[0]->height(),bottom[0]->width() };
+  vector<int> shape_counter(arr_counter,arr_counter+sizeof(arr_counter)/sizeof(int));
+  counter_.Reshape(shape_counter);
+#else
   top[0]->Reshape({ bottom[0]->num(), 5, bottom[0]->height(), bottom[0]->width() });
   if (output_vector_) {
     top[1]->Reshape({ bottom[0]->num(), 1, 5 });//will be modified on the fly.
   }
   counter_.Reshape({ bottom[0]->num(),1,bottom[0]->height(),bottom[0]->width() });
+#endif
 }
 
 template <typename Dtype>
@@ -111,7 +124,13 @@ void PredictBoxLayer<Dtype>::Forward_cpu(
 
   if (output_vector_) {
     if (num == 1 && count > 0) {
+#if __cplusplus < 201103L
+      int arr[] = { bottom[0]->num(), count, 5 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ bottom[0]->num(), count, 5 });
+#endif
       int i = 0;
       for (int x = 0; x < output_width; x++) {
         for (int y = 0; y < output_height; y++) {
@@ -127,7 +146,13 @@ void PredictBoxLayer<Dtype>::Forward_cpu(
       }
     }
     else {
+#if __cplusplus < 201103L
+      int arr[] = { bottom[0]->num(), 1, 5 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ bottom[0]->num(), 1, 5 });
+#endif
       caffe_set<Dtype>(top[1]->count(), 0, top[1]->mutable_cpu_data());
     }
   }
diff --git a/src/caffe/layers/predict_box_layer.cu b/src/caffe/layers/predict_box_layer.cu
index e28f123ad..87d74bd4e 100644
--- a/src/caffe/layers/predict_box_layer.cu
+++ b/src/caffe/layers/predict_box_layer.cu
@@ -139,7 +139,13 @@ void PredictBoxLayer<Dtype>::Forward_gpu(
     const Dtype* score_data_cpu = bottom[0]->cpu_data();
     const Dtype* bb_data_cpu = top[0]->cpu_data();
     if (num == 1 && count > 0) {
+#if __cplusplus < 201103L
+      int arr[] = { bottom[0]->num(), (int)count, 5 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ bottom[0]->num(), (int)count, 5 });
+#endif
       int i = 0;
       for (int x = 0; x < output_width; x++) {
         for (int y = 0; y < output_height; y++) {
@@ -155,7 +161,13 @@ void PredictBoxLayer<Dtype>::Forward_gpu(
       }
     }
     else {
+#if __cplusplus < 201103L
+      int arr[] = { bottom[0]->num(), 1, 5 };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      top[1]->Reshape(shape);
+#else
       top[1]->Reshape({ bottom[0]->num(), 1, 5 });
+#endif
       caffe_gpu_set<Dtype>(top[1]->count(), 0, top[1]->mutable_gpu_data());
     }
   }
diff --git a/src/caffe/layers/soft_contrastive_loss_layer.cpp b/src/caffe/layers/soft_contrastive_loss_layer.cpp
index 43688544e..f96bc8c03 100644
--- a/src/caffe/layers/soft_contrastive_loss_layer.cpp
+++ b/src/caffe/layers/soft_contrastive_loss_layer.cpp
@@ -27,9 +27,19 @@ void SoftContrastiveLossLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom
   LossLayer<Dtype>::Reshape(bottom, top);
   if (top.size() >= 2) {
     // positive distance, negative distance.
+#if __cplusplus < 201103L
+    int arr[] = { 2 };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[1]->Reshape(shape);
+  }
+  int arr_sum_exp[] = { bottom[0]->num(), 1 };
+  vector<int> shape_sum_exp(arr_sum_exp,arr_sum_exp+sizeof(arr_sum_exp)/sizeof(int));
+  sum_exp_.Reshape(shape_sum_exp);
+#else
     top[1]->Reshape({ 2 });
   }
   sum_exp_.Reshape({ bottom[0]->num(), 1 });
+#endif
 }
 
 template <typename Dtype>
@@ -42,7 +52,7 @@ void SoftContrastiveLossLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bo
   int num = bottom[0]->num();
   int count = bottom[0]->count();
   int dim = count / num;
-  Dtype weighted_count = num * (abs(positive_weight_) + (dim - 1)*abs(negative_weight_));
+//  Dtype weighted_count = num * (abs(positive_weight_) + (dim - 1)*abs(negative_weight_));
   Dtype positive_distance = Dtype(0);
   Dtype negative_distance = Dtype(0);
   Dtype* loss = top[0]->mutable_cpu_data();
@@ -110,11 +120,11 @@ void SoftContrastiveLossLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& t
     int count = bottom[0]->count();
     int dim = count / num;
 
-    Dtype negative_sum = Dtype(0);
+ //   Dtype negative_sum = Dtype(0);
 
     for (int i = 0; i < num; ++i) {
-      Dtype min_negative_distance = FLT_MAX;
-      int min_negative_index = 0;
+//      Dtype min_negative_distance = FLT_MAX;
+//      int min_negative_index = 0;
       for (int j = 0; j < dim; ++j) {
         if (j == static_cast<int>(label[i])) {
           if (bottom_data[i * dim + j] > positive_margin_ && bottom_data[i * dim + j] < positive_outlier_thresh_) {
diff --git a/src/caffe/layers/softmax_loss_layer.cpp b/src/caffe/layers/softmax_loss_layer.cpp
index 21576170b..581613152 100644
--- a/src/caffe/layers/softmax_loss_layer.cpp
+++ b/src/caffe/layers/softmax_loss_layer.cpp
@@ -64,7 +64,13 @@ void SoftmaxWithLossLayer<Dtype>::LayerSetUp(
   softmax_axis_ =
     bottom[0]->CanonicalAxisIndex(this->layer_param_.softmax_param().axis());
   if (has_class_weight_) {
+#if __cplusplus < 201103L
+    int arr[] = { bottom[0]->shape(softmax_axis_) };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    class_weight_.Reshape(shape);
+#else
     class_weight_.Reshape({ bottom[0]->shape(softmax_axis_) });
+#endif
     CHECK_EQ(this->layer_param_.softmax_param().class_weight().size(), bottom[0]->shape(softmax_axis_));
     for (int i = 0; i < bottom[0]->shape(softmax_axis_); i++) {
       class_weight_.mutable_cpu_data()[i] = (Dtype)this->layer_param_.softmax_param().class_weight(i);
@@ -72,7 +78,13 @@ void SoftmaxWithLossLayer<Dtype>::LayerSetUp(
   }
   else {
     if (bottom.size() == 3) {
+#if __cplusplus < 201103L
+      int arr[] = { bottom[0]->shape(softmax_axis_) };
+      vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+      class_weight_.Reshape(shape);
+#else
       class_weight_.Reshape({ bottom[0]->shape(softmax_axis_) });
+#endif
       for (int i = 0; i < bottom[0]->shape(softmax_axis_); i++) {
         class_weight_.mutable_cpu_data()[i] = (Dtype)1.0;
       }
@@ -97,8 +109,15 @@ void SoftmaxWithLossLayer<Dtype>::Reshape(
       bottom[0]->CanonicalAxisIndex(this->layer_param_.softmax_param().axis());
   outer_num_ = bottom[0]->count(0, softmax_axis_);
   inner_num_ = bottom[0]->count(softmax_axis_ + 1);
+#if __cplusplus < 201103L
+  int arr[] = { outer_num_, inner_num_ };
+  vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+  counts_.Reshape(shape);
+  loss_.Reshape(shape);
+#else
   counts_.Reshape({ outer_num_, inner_num_ });
   loss_.Reshape({ outer_num_, inner_num_ });
+#endif
   CHECK_EQ(outer_num_ * inner_num_, bottom[1]->count())
       << "Number of labels must match number of predictions; "
       << "e.g., if softmax axis == 1 and prediction shape is (N, C, H, W), "
diff --git a/src/caffe/layers/subregion_layer.cpp b/src/caffe/layers/subregion_layer.cpp
index 299079a47..0d01ffbd2 100644
--- a/src/caffe/layers/subregion_layer.cpp
+++ b/src/caffe/layers/subregion_layer.cpp
@@ -28,12 +28,25 @@ void SubRegionLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
       << "corresponding to (num, channels, height, width)";
   CHECK_EQ(bottom[1]->channels() % 2, 0) << "The coordinate blob's size must be able to divided by 2!";
   int num_point = bottom[1]->channels() / 2;
+#if __cplusplus < 201103L
+  if (as_dim_ == 0) {
+    int arr[] = { bottom[0]->num() * num_point, bottom[0]->channels(), height_, width_ };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[0]->Reshape(shape);
+  }
+  else {
+    int arr[] = { bottom[0]->num(), bottom[0]->channels() * num_point, height_, width_ };
+    vector<int> shape(arr,arr+sizeof(arr)/sizeof(int));
+    top[0]->Reshape(shape);
+  }
+#else
   if (as_dim_ == 0) {
     top[0]->Reshape({ bottom[0]->num() * num_point, bottom[0]->channels(), height_, width_ });
   }
   else {
     top[0]->Reshape({ bottom[0]->num(), bottom[0]->channels() * num_point, height_, width_ });
   }
+#endif
   if (top.size() == 3) {
     top[1]->ReshapeLike(*bottom[1]);
     top[2]->ReshapeLike(*bottom[1]);
diff --git a/src/caffe/layers/transformer_layer.cpp b/src/caffe/layers/transformer_layer.cpp
index 1ba9b78b7..034ba1e5b 100644
--- a/src/caffe/layers/transformer_layer.cpp
+++ b/src/caffe/layers/transformer_layer.cpp
@@ -34,7 +34,7 @@ namespace caffe {
   template <typename Dtype>
   void TransformerLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
     const vector<Blob<Dtype>*>& top) {
-    const Dtype* bottom_data = bottom[0]->cpu_data();
+//    const Dtype* bottom_data = bottom[0]->cpu_data();
     Dtype* top_data = top[0]->mutable_cpu_data();
     const Dtype* theta_data = bottom[1]->cpu_data();
     const Dtype* CoordinateTarget_data = CoordinateTarget.cpu_data();
@@ -94,7 +94,7 @@ namespace caffe {
     const vector<bool>& propagate_down,
     const vector<Blob<Dtype>*>& bottom) {
     const Dtype* top_diff = top[0]->cpu_diff();
-    const Dtype* top_data = top[0]->cpu_data();
+ //   const Dtype* top_data = top[0]->cpu_data();
     Dtype* data_diff = bottom[0]->mutable_cpu_diff();
     Dtype* theta_diff = bottom[1]->mutable_cpu_diff();
     int num = bottom[0]->shape(0);
diff --git a/windows/caffe.binding/CMakeLists.txt b/windows/caffe.binding/CMakeLists.txt
new file mode 100644
index 000000000..086d18afd
--- /dev/null
+++ b/windows/caffe.binding/CMakeLists.txt
@@ -0,0 +1,28 @@
+if(NOT HAVE_BINGING)
+  message(STATUS "caffe binding interface is disabled. Building without it...")
+  return()
+endif()
+
+# ---[ caffe_binding project
+project(caffe_binding C CXX)
+
+file(GLOB_RECURSE caffe_binding_srcs ${PROJECT_SOURCE_DIR}/*.cpp)
+
+SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
+
+add_library(caffe_binding SHARED ${caffe_binding_srcs})
+caffe_default_properties(caffe_binding)
+set_target_properties(caffe_binding PROPERTIES PREFIX "" OUTPUT_NAME "caffe_binding")
+target_link_libraries(caffe_binding PUBLIC ${Caffe_LINK})
+
+if(UNIX OR APPLE)
+    set(__linkname "${PROJECT_SOURCE_DIR}/caffe_binding.so")
+    add_custom_command(TARGET caffe_binding POST_BUILD
+                       COMMAND ln -sf $<TARGET_LINKER_FILE:caffe_binding> "${__linkname}"
+                       COMMENT "Creating symlink ${__linkname} -> ${PROJECT_BINARY_DIR}/lib/caffe_binding${Caffe_POSTFIX}.so")
+endif()
+
+# ---[ Install
+# caffe_binding.so
+install(TARGETS caffe_binding  DESTINATION caffe_binding)
+
diff --git a/windows/caffe.binding/CaffeBinding.cpp b/windows/caffe.binding/CaffeBinding.cpp
index 496d72613..b7bea1072 100644
--- a/windows/caffe.binding/CaffeBinding.cpp
+++ b/windows/caffe.binding/CaffeBinding.cpp
@@ -1,8 +1,12 @@
 #include "CaffeBinding.h"
-#include <caffe\caffe.hpp>
-#include <caffe\layers\memory_data_layer.hpp>
+#include <caffe/caffe.hpp>
+#include <caffe/layers/memory_data_layer.hpp>
 #include <boost/thread.hpp>
 
+#if __cplusplus < 201402L
+#include <boost/make_unique.hpp>
+#endif
+
 using namespace caffe;
 using namespace std;
 
@@ -26,8 +30,13 @@ int CaffeBinding::AddNet(string model_definition, string weights, int gpu_id) {
 
 std::unordered_map<std::string, DataBlob> CaffeBinding::Forward(int net_id) {
   if (!(*predictors_[net_id]).get()) {
+#if __cplusplus >= 201402L
     auto predictor =
       std::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#else
+    auto predictor =
+      boost::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#endif
     predictor->ShareTrainedLayersWith(nets_[net_id]);
     (*predictors_[net_id]).reset(predictor.release());
   }
@@ -48,8 +57,13 @@ std::unordered_map<std::string, DataBlob> CaffeBinding::Forward(std::vector<cv::
 
 void CaffeBinding::SetMemoryDataLayer(std::string layer_name, std::vector<cv::Mat>&& input_image, int net_id) {
   if (!(*predictors_[net_id]).get()) {
+#if __cplusplus >= 201402L
     auto predictor =
       std::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#else
+    auto predictor =
+      boost::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#endif
     predictor->ShareTrainedLayersWith(nets_[net_id]);
     (*predictors_[net_id]).reset(predictor.release());
   }
@@ -62,8 +76,13 @@ void CaffeBinding::SetMemoryDataLayer(std::string layer_name, std::vector<cv::Ma
 
 void CaffeBinding::SetBlobData(std::string blob_name, std::vector<int> blob_shape, float* data, int net_id) {
   if (!(*predictors_[net_id]).get()) {
+#if __cplusplus >= 201402L
     auto predictor =
       std::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#else
+    auto predictor =
+      boost::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+#endif
     predictor->ShareTrainedLayersWith(nets_[net_id]);
     (*predictors_[net_id]).reset(predictor.release());
   }
@@ -74,8 +93,13 @@ void CaffeBinding::SetBlobData(std::string blob_name, std::vector<int> blob_shap
 
 DataBlob CaffeBinding::GetBlobData(std::string blob_name, int net_id) {
   if (!(*predictors_[net_id]).get()) {
+#if __cplusplus >= 201402L
     auto predictor =
-      std::make_unique<caffe::Net<float>>(prototxts[net_id], Phase::TEST);
+      std::make_unique<caffe::Net<float> >(prototxts[net_id], Phase::TEST);
+#else
+    auto predictor =
+      boost::make_unique<caffe::Net<float> >(prototxts[net_id], Phase::TEST);
+#endif
     predictor->ShareTrainedLayersWith(nets_[net_id]);
     (*predictors_[net_id]).reset(predictor.release());
   }
diff --git a/windows/caffe.binding/CaffeBinding.h b/windows/caffe.binding/CaffeBinding.h
index 10d5a8e81..8338c367b 100644
--- a/windows/caffe.binding/CaffeBinding.h
+++ b/windows/caffe.binding/CaffeBinding.h
@@ -1,12 +1,16 @@
 #pragma once
 
+#ifdef _MSC_VER
 #ifdef CAFFEBINDING_EXPORTS
 #define CAFFE_DLL __declspec(dllexport)
 #else
 #define CAFFE_DLL __declspec(dllimport)
 #endif
+#else
+#define CAFFE_DLL
+#endif
 
-#include <opencv2\opencv.hpp>
+#include <opencv2/opencv.hpp>
 #include <string>
 #include <unordered_map>
 #include <vector>
@@ -35,4 +39,4 @@ namespace caffe {
     void SetDevice(int gpu_id);
     ~CaffeBinding();
   };
-}
\ No newline at end of file
+}