Resolve reference counting problem in Torch importer.

modules/dnn/src/torch/torch_importer.cpp

@@ -185,9 +185,6 @@ struct TorchImporter : public ::cv::dnn::Importer
 
     void readTorchStorage(int index, int type = -1)
     {
-        if (storages.count(index))
-            return;
-
         long size = readLong();
         Mat storageMat(1, size, (type != CV_USRTYPE1) ? type : CV_64F); //handle LongStorage as CV_64F Mat
 
@@ -225,7 +222,6 @@ struct TorchImporter : public ::cv::dnn::Importer
         }
 
         storages.insert(std::make_pair(index, storageMat));
-        readedIndexes.insert(index);
     }
 
     void readTorchTable(Dict &scalarParams, std::map<String, Blob> &tensorParams)
@@ -261,10 +257,7 @@ struct TorchImporter : public ::cv::dnn::Importer
             if (vtype == TYPE_TORCH)
             {
                 int index = readInt();
-                if (tensors.count(index) == 0)
-                {
                 readTorchObject(index);
-                }
 
                 if (tensors.count(index)) //tensor was readed
                 {
@@ -311,9 +304,6 @@ struct TorchImporter : public ::cv::dnn::Importer
 
     void readTorchTensor(int indexTensor, int typeTensor)
     {
-        if (tensors.count(indexTensor))
-            return;
-
         int ndims = readInt();
         AutoBuffer<long, 4> sizes(ndims);
         AutoBuffer<long, 4> steps(ndims);
@@ -394,6 +384,14 @@ struct TorchImporter : public ::cv::dnn::Importer
 
     void readTorchObject(int index)
     {
+        if(readedIndexes.count(index))
+        {
+            if(!storages.count(index) && !tensors.count(index))
+                CV_Error(Error::StsNotImplemented, "Objects which have multiple references are not supported");
+            else
+                return;
+        }
+
         String className = readTorchClassName();
         String nnName;
         std::cout << "Class: " << className << std::endl;
@@ -409,8 +407,6 @@ struct TorchImporter : public ::cv::dnn::Importer
         }
         else if (isNNClass(className, nnName))
         {
-            CV_Assert(!readedIndexes.count(index));
-
             Dict scalarParams;
             std::map<String, Blob> tensorParams;
 
@@ -524,6 +520,8 @@ struct TorchImporter : public ::cv::dnn::Importer
         {
             CV_Error(Error::StsNotImplemented, "Unsupported Torch class \"" + className + "\"");
         }
+
+        readedIndexes.insert(index);
     }
 
     void readObject()
@@ -533,13 +531,9 @@ struct TorchImporter : public ::cv::dnn::Importer
         if (typeidx == TYPE_TORCH)
         {
             int index = readInt();
-
-            if (readedIndexes.count(index) == 0)
-            {
             readTorchObject(index);
             readedIndexes.insert(index);
         }
-        }
         else if (typeidx == TYPE_NIL)
             return;
         else if (typeidx == TYPE_NUMBER)

modules/dnn/testdata/dnn/torch/torch_gen_test_data.lua

@@ -27,7 +27,7 @@ function save(net, input, label)
 end
 
 local net_simple = nn.Sequential()
---net_simple:add(nn.ReLU())
+net_simple:add(nn.ReLU())
 net_simple:add(nn.SpatialConvolution(3,64, 11,7, 3,4, 3,2))
 net_simple:add(nn.SpatialMaxPooling(4,5, 3,2, 1,2))
 net_simple:add(nn.Sigmoid())