Cleaned up the tensor_conv interface a little. Also fixed an error in the spec

for this object.

dlib/dnn/cudnn_dlibapi.cpp

@@ -265,10 +265,6 @@ namespace dlib
         ) : 
             filter_handle(nullptr),
             conv_handle(nullptr),
-            out_num_samples(0),
-            out_k(0),
-            out_nr(0),
-            out_nc(0),
             forward_algo(0),
             forward_workspace_size_in_bytes(0),
             forward_workspace(nullptr),
@@ -279,6 +275,7 @@ namespace dlib
             backward_filters_workspace_size_in_bytes(0),
             backward_filters_workspace(nullptr)
         {
+            clear();
         }
 
         void tensor_conv::
@@ -313,6 +310,17 @@ namespace dlib
             backward_filters_workspace = nullptr;
             backward_filters_algo = 0;
             backward_filters_workspace_size_in_bytes = 0;
+
+            stride_y = 0;
+            stride_x = 0;
+            data_num_samples = 0;
+            data_k = 0;
+            data_nr = 0;
+            data_nc = 0;
+            filters_num_samples = 0;
+            filters_k = 0;
+            filters_nr = 0;
+            filters_nc = 0;
         }
 
         void tensor_conv::
@@ -324,11 +332,36 @@ namespace dlib
         ) 
         {
             DLIB_CASSERT(data.k() == filters.k(),"");
+
+            // if the last call to setup gave the same exact settings then don't do
+            // anything.
+            if (stride_y_ == stride_y && 
+                stride_x_ == stride_x &&
+                data_num_samples == data.num_samples() &&
+                data_k == data.k() &&
+                data_nr == data.nr() &&
+                data_nc == data.nc() &&
+                filters_num_samples == filters.num_samples() &&
+                filters_k == filters.k() &&
+                filters_nr == filters.nr() &&
+                filters_nc == filters.nc())
+            {
+                return;
+            }
+
             clear();
             try
             {
                 stride_y = stride_y_;
                 stride_x = stride_x_;
+                data_num_samples = data.num_samples();
+                data_k = data.k();
+                data_nr = data.nr();
+                data_nc = data.nc();
+                filters_num_samples = filters.num_samples();
+                filters_k = filters.k();
+                filters_nr = filters.nr();
+                filters_nc = filters.nc();
 
                 CHECK_CUDNN(cudnnCreateFilterDescriptor((cudnnFilterDescriptor_t*)&filter_handle));
                 CHECK_CUDNN(cudnnSetFilter4dDescriptor((cudnnFilterDescriptor_t)filter_handle, 
@@ -446,18 +479,22 @@ namespace dlib
         void tensor_conv::operator() (
             resizable_tensor& output,
             const tensor& data,
-            const tensor& filters
+            const tensor& filters,
+            int stride_y,
+            int stride_x
         )
         {
             DLIB_ASSERT(is_same_object(output,data) == false,"");
             DLIB_ASSERT(is_same_object(output,filters) == false,"");
 
+            setup(data,filters,stride_y,stride_x);
+
             output.set_size(out_num_samples, out_k, out_nr, out_nc);
 
             DLIB_ASSERT(output.num_samples() == data.num_samples(),out_num_samples << "  " << data.num_samples());
             DLIB_ASSERT(output.k() == filters.num_samples(),"");
-            DLIB_ASSERT(output.nr() == 1+(data.nr()-1)/stride_y,"");
-            DLIB_ASSERT(output.nc() == 1+(data.nc()-1)/stride_x,"");
+            DLIB_ASSERT(output.nr() == 1+(data.nr()-filters.nr()%2)/stride_y,"");
+            DLIB_ASSERT(output.nc() == 1+(data.nc()-filters.nc()%2)/stride_x,output.nc() << "  " <<1+(data.nc()-1)/stride_x << " : " << data.nc() << "  " << stride_x);
 
             const float alpha = 1;
             const float beta = 0;

dlib/dnn/cudnn_dlibapi.h

@@ -144,31 +144,20 @@ namespace dlib
             void clear(
             );
 
-            void setup(
-                const tensor& data,
-                const tensor& filters,
-                int stride_y,
-                int stride_x
-            );
-            /*!
-                requires
-                    - filters.k() == data.k()
-                    - stride_y > 0
-                    - stride_x > 0
-            !*/
-
             ~tensor_conv (
             );
 
             void operator() (
                 resizable_tensor& output,
                 const tensor& data,
-                const tensor& filters
+                const tensor& filters,
+                int stride_y,
+                int stride_x
             );
             /*!
                 requires
-                    - The dimensions of data and filters are the same as the ones given 
-                      to the last call to setup().
+                    - stride_y > 0
+                    - stride_x > 0
                     - is_same_object(output,data) == false
                     - is_same_object(output,filters) == false
                 ensures
@@ -176,8 +165,8 @@ namespace dlib
                     - filters contains filters.num_samples() filters. 
                     - #output.num_samples() == data.num_samples()
                     - #output.k() == filters.num_samples()
-                    - #output.nr() == 1+(data.nr()-1)/stride_y
-                    - #output.nc() == 1+(data.nc()-1)/stride_x
+                    - #output.nr() == 1+(data.nr()-filters.nr()%2)/stride_y
+                    - #output.nc() == 1+(data.nc()-filters.nc()%2)/stride_x
             !*/
 
             void get_gradient_for_data (
@@ -188,9 +177,9 @@ namespace dlib
             /*!
                 requires
                     - filters has the same dimensions as the filters object give to the 
-                      last call to setup().
+                      last call to operator().
                     - data_gradient has the same dimensions as the data object give to the
-                      last call to setup().
+                      last call to operator().
                     - gradient_input has the same dimensions as the output of operator().
                     - is_same_object(data_gradient,filters) == false
                     - is_same_object(data_gradient,gradient_input) == false
@@ -209,9 +198,9 @@ namespace dlib
             /*!
                 requires
                     - filters_gradient has the same dimensions as the filters object give
-                      to the last call to setup().
+                      to the last call to operator().
                     - data has the same dimensions as the data object give to the last call
-                      to setup().
+                      to operator().
                     - gradient_input has the same dimensions as the output of operator().
                     - is_same_object(filters_gradient,data) == false
                     - is_same_object(filters_gradient,gradient_input) == false
@@ -223,10 +212,29 @@ namespace dlib
             !*/
 
         private:
-            void* filter_handle;
-            void* conv_handle;
+
+            void setup(
+                const tensor& data,
+                const tensor& filters,
+                int stride_y,
+                int stride_x
+            );
+            /*!
+                requires
+                    - filters.k() == data.k()
+                    - stride_y > 0
+                    - stride_x > 0
+            !*/
+
+            // These variables record the type of data given to the last call to setup().
             int stride_y;
             int stride_x;
+            long data_num_samples, data_k, data_nr, data_nc;
+            long filters_num_samples, filters_k, filters_nr, filters_nc;
+
+
+            void* filter_handle;
+            void* conv_handle;
 
             // dimensions of the output tensor from operator()
             int out_num_samples;

dlib/dnn/tensor_tools.cpp

@@ -277,7 +277,8 @@ namespace dlib { namespace tt
     }
 
     void tensor_conv::
-    setup(
+    operator() (
+        resizable_tensor& output,
         const tensor& data,
         const tensor& filters,
         int stride_y,
@@ -285,22 +286,7 @@ namespace dlib { namespace tt
     )
     {
 #ifdef DLIB_USE_CUDA
-        impl.setup(data, filters, stride_y, stride_x);
-#else
-        // TODO
-        DLIB_CASSERT(false,"");
-#endif
-    }
-
-    void tensor_conv::
-    operator() (
-        resizable_tensor& output,
-        const tensor& data,
-        const tensor& filters
-    )
-    {
-#ifdef DLIB_USE_CUDA
-        impl(output, data, filters);
+        impl(output, data, filters, stride_y, stride_x);
 #else
         // TODO
         DLIB_CASSERT(false,"");

dlib/dnn/tensor_tools.h

@@ -397,7 +397,8 @@ namespace dlib { namespace tt
         void clear(
         );
 
-        void setup(
+        void operator() (
+            resizable_tensor& output,
             const tensor& data,
             const tensor& filters,
             int stride_y,
@@ -405,20 +406,8 @@ namespace dlib { namespace tt
         );
         /*!
             requires
-                - filters.k() == data.k()
                 - stride_y > 0
                 - stride_x > 0
-        !*/
-
-        void operator() (
-            resizable_tensor& output,
-            const tensor& data,
-            const tensor& filters
-        );
-        /*!
-            requires
-                - The dimensions of data and filters are the same as the ones given 
-                  to the last call to setup().
                 - is_same_object(output,data) == false
                 - is_same_object(output,filters) == false
             ensures
@@ -426,8 +415,8 @@ namespace dlib { namespace tt
                     - filters contains filters.num_samples() filters. 
                     - #output.num_samples() == data.num_samples()
                     - #output.k() == filters.num_samples()
-                    - #output.nr() == 1+(data.nr()-1)/stride_y
-                    - #output.nc() == 1+(data.nc()-1)/stride_x
+                    - #output.nr() == 1+(data.nr()-filters.nr()%2)/stride_y
+                    - #output.nc() == 1+(data.nc()-filters.nc()%2)/stride_x
         !*/
 
         void get_gradient_for_data (
@@ -438,9 +427,9 @@ namespace dlib { namespace tt
         /*!
             requires
                 - filters has the same dimensions as the filters object give to the last
-                  call to setup().
+                  call to operator().
                 - data_gradient has the same dimensions as the data object give to the last
-                  call to setup().
+                  call to operator().
                 - gradient_input has the same dimensions as the output of operator().
                 - is_same_object(data_gradient,filters) == false
                 - is_same_object(data_gradient,gradient_input) == false
@@ -459,9 +448,9 @@ namespace dlib { namespace tt
         /*!
             requires
                 - filters_gradient has the same dimensions as the filters object give to
-                  the last call to setup().
+                  the last call to operator().
                 - data has the same dimensions as the data object give to the last call to
-                  setup().
+                  operator().
                 - gradient_input has the same dimensions as the output of operator().
                 - is_same_object(filters_gradient,data) == false
                 - is_same_object(filters_gradient,gradient_input) == false