1. Enable network state write/read

[ti-machine-learning/ti-machine-learning.git] / src / app / cnn / scene / appCNNSceneClassifyOpenMP.c

diff --git a/src/app/cnn/scene/appCNNSceneClassifyOpenMP.c b/src/app/cnn/scene/appCNNSceneClassifyOpenMP.c
index cb8923cbffc88bef59ef78ce28c5edcd56f74457..a920b34ab2ec641135021033f0c0a6b0573146e9 100644 (file)
--- a/src/app/cnn/scene/appCNNSceneClassifyOpenMP.c
+++ b/src/app/cnn/scene/appCNNSceneClassifyOpenMP.c
@@ -114,7 +114,7 @@ int appCNNSceneClassifyOpenMP(timlConvNeuralNetwork **cnnTeam, int teamNum, floa
       }
    }
 
-   #pragma omp parallel num_threads(teamNum) private(threadIndex, m, k, i, j, p)
+   #pragma omp parallel num_threads(teamNum) private(threadIndex, m, k, i, j, p, rowStart, rowEnd, colStart, colEnd)
    {
       #pragma omp for collapse(2)
 
@@ -142,10 +142,16 @@ int appCNNSceneClassifyOpenMP(timlConvNeuralNetwork **cnnTeam, int teamNum, floa
             }
 
             // cnn Forward Propagation
-            err = timlCNNForwardPropagation(cnnTeam[threadIndex], paddedImage + paddedDim*threadIndex, paddedDim);
+            err = timlCNNLoadImage(cnnTeam[threadIndex], paddedImage + paddedDim*threadIndex, paddedRow, paddedCol, channel, 1);
+            err = timlCNNForwardPropagation(cnnTeam[threadIndex]);
 
             // labeling
-            appCNNSceneLabelMatrix(cnnTeam[threadIndex]->tail->featureMap, cnnTeam[threadIndex]->tail->row, cnnTeam[threadIndex]->tail->col, cnnTeam[threadIndex]->tail->channel, m, k, labelMatrix, row, col);
+            for (i = 0; i < cnnTeam[threadIndex]->tail->row; i++) {
+               for (j = 0; j < cnnTeam[threadIndex]->tail->col; j++) {
+                  labelMatrix[k + resolutionLossCol/2 + j*resolutionLossCol + (m + resolutionLossRow/2 + i*resolutionLossRow)*col] = cnnTeam[threadIndex]->tail->accuracyParams.label[i*cnnTeam[threadIndex]->tail->col + j];
+               }
+            }
+
          } // end of k loop
       } // end of m loop
    } // end of OpenMP