diff --git a/modules/gpu/src/constantspacebp_gpu.cpp b/modules/gpu/src/constantspacebp_gpu.cpp
index b6842b7a8..09b8e89ca 100644
--- a/modules/gpu/src/constantspacebp_gpu.cpp
+++ b/modules/gpu/src/constantspacebp_gpu.cpp
@@ -185,9 +185,9 @@ static void stereo_csbp_gpu_operator(int& ndisp, int& iters, int& levels, int& n
     step_pyr[0] = data_cost.step / type_size;
 
     Size temp_size = data_cost_size;
-    if (data_cost.step * data_cost_size.height < static_cast<size_t>(step_pyr[levels - 1]) * rows_pyr[levels - 1] * ndisp)
+    if (data_cost_size.width * data_cost_size.height < static_cast<size_t>(step_pyr[levels - 1]) * rows_pyr[levels - 1] * ndisp)
     {
-        temp_size = Size(step_pyr[levels - 1], rows_pyr[levels - 1] * nr_plane);
+        temp_size = Size(step_pyr[levels - 1], rows_pyr[levels - 1] * ndisp);
     }
 
     temp1.create(temp_size, msg_type);
diff --git a/modules/gpu/src/cuda/constantspacebp.cu b/modules/gpu/src/cuda/constantspacebp.cu
index 3703afc4a..a94d374d7 100644
--- a/modules/gpu/src/cuda/constantspacebp.cu
+++ b/modules/gpu/src/cuda/constantspacebp.cu
@@ -269,7 +269,7 @@ namespace cv { namespace gpu { namespace csbp
         const int threadsNum = 256;
         const size_t smem_size = threadsNum * sizeof(float);
         
-        dim3 threads(winsz, 1, threadsNum/winsz);
+        dim3 threads(winsz, 1, threadsNum / winsz);
         dim3 grid(w, h, 1);        
         grid.y *= divUp(ndisp, threads.z);
         
@@ -278,7 +278,7 @@ namespace cv { namespace gpu { namespace csbp
         case 1: csbp_kernels::data_init<T, winsz, 1><<<grid, threads, smem_size, stream>>>(level, rows, cols, h); break;
         case 3: csbp_kernels::data_init<T, winsz, 3><<<grid, threads, smem_size, stream>>>(level, rows, cols, h); break;
         default: cv::gpu::error("Unsupported channels count", __FILE__, __LINE__);
-        }            
+        }
     }
       
     typedef void (*DataInitCaller)(int cols, int rows, int w, int h, int level, int ndisp, int channels, const cudaStream_t& stream);
@@ -419,7 +419,7 @@ namespace cv { namespace gpu { namespace csbp
       
     typedef void (*ComputeDataCostCaller)(const DevMem2D& disp_selected_pyr, const DevMem2D& data_cost, 
                            int h, int w, int level, int nr_plane, int channels, const cudaStream_t& stream);
-
+    
     void compute_data_cost(const DevMem2D& disp_selected_pyr, const DevMem2D& data_cost, size_t msg_step1, size_t msg_step2, int msg_type,
                            int h, int w, int h2, int level, int nr_plane, int channels, const cudaStream_t& stream)
     {
diff --git a/tests/gpu/CMakeLists.txt b/tests/gpu/CMakeLists.txt
index d20c65020..f73969353 100644
--- a/tests/gpu/CMakeLists.txt
+++ b/tests/gpu/CMakeLists.txt
@@ -22,6 +22,7 @@ include_directories	(
 			"${CMAKE_SOURCE_DIR}/modules/legacy/include"
 			"${CMAKE_SOURCE_DIR}/modules/contrib/include"
 			"${CMAKE_SOURCE_DIR}/modules/gpu/include"
+			"${CMAKE_SOURCE_DIR}/modules/ml/include"
 			"${CMAKE_CURRENT_SOURCE_DIR}/src"
 			"${CMAKE_CURRENT_BINARY_DIR}"
 			)