optimize yolov3 output extraction speed. (#1398)

src/layer/yolov3detectionoutput.cpp

@@ -221,16 +221,17 @@ int Yolov3DetectionOutput::forward(const std::vector<Mat>& bottom_blobs, std::ve
 
                     // find class index with max class score
                     int class_index = 0;
-                    float class_score = 0.f;
+                    float class_score = -std::numeric_limits<float>::max();
                     for (int q = 0; q < num_class; q++)
                     {
-                        float score = sigmoid(scores.channel(q).row(i)[j]);
+                        float score = scores.channel(q).row(i)[j];
                         if (score > class_score)
                         {
                             class_index = q;
                             class_score = score;
                         }
                     }
+                    class_score = sigmoid(class_score);
 
                     //printf( "%d %f %f\n", class_index, box_score, class_score);