Fix Darknet eltwise

modules/dnn/src/darknet/darknet_io.cpp

@@ -420,7 +420,7 @@ namespace cv {
                     if (alpha != 1)
                     {
                         std::vector<float> coeffs(2, 1);
-                        coeffs[1] = alpha;
+                        coeffs[0] = alpha;
                         shortcut_param.set("coeff", DictValue::arrayReal<float*>(&coeffs[0], coeffs.size()));
                     }
 
@@ -431,8 +431,8 @@ namespace cv {
                     lp.layer_name = layer_name;
                     lp.layer_type = shortcut_param.type;
                     lp.layerParams = shortcut_param;
-                    lp.bottom_indexes.push_back(fused_layer_names.at(from));
                     lp.bottom_indexes.push_back(last_layer);
+                    lp.bottom_indexes.push_back(fused_layer_names.at(from));
                     last_layer = layer_name;
                     net->layers.push_back(lp);
 
@@ -599,7 +599,7 @@ namespace cv {
 
                         current_channels = 0;
                         for (size_t k = 0; k < layers_vec.size(); ++k) {
-                            layers_vec[k] = layers_vec[k] > 0 ? layers_vec[k] : (layers_vec[k] + layers_counter);
+                            layers_vec[k] = layers_vec[k] >= 0 ? layers_vec[k] : (layers_vec[k] + layers_counter);
                             current_channels += net->out_channels_vec[layers_vec[k]];
                         }
 
@@ -645,8 +645,6 @@ namespace cv {
                         int from = std::atoi(bottom_layer.c_str());
 
                         from = from < 0 ? from + layers_counter : from;
-                        current_channels = net->out_channels_vec[from];
-
                         setParams.setShortcut(from, alpha);
                     }
                     else if (layer_type == "upsample")

modules/dnn/src/layers/eltwise_layer.cpp

@@ -114,11 +114,11 @@ public:
         CV_Assert(op == SUM || coeffs.size() == 0);
 
         int dims = inputs[0].size();
+        // Number of channels in output shape is determined by the first input tensor.
         int numChannels = inputs[0][1];
         for (int i = 1; i < inputs.size(); i++)
         {
             CV_Assert(inputs[0][0] == inputs[i][0]);
-            numChannels = std::max(numChannels, inputs[i][1]);
 
             // It's allowed for channels axis to be different.
             for (int j = 2; j < dims; j++)
@@ -177,7 +177,7 @@ public:
                 CV_Assert(srcs[i].type() == dst.type() &&
                           srcs[i].isContinuous());
                 // Sort srcs and coefficients in the order by number of channels
-                for( int j = i - 1; j >= 1 && p.srcs[j - 1]->size[1] < p.srcs[j]->size[1]; j++ )
+                for( int j = i; j >= 1 && p.srcs[j - 1]->size[1] < p.srcs[j]->size[1]; j-- )
                 {
                     std::swap(p.srcs[j - 1], p.srcs[j]);
                     if (!p.coeffs.empty())

modules/dnn/test/test_darknet_importer.cpp

@@ -445,6 +445,8 @@ TEST_P(Test_Darknet_layers, shortcut)
 {
     testDarknetLayer("shortcut");
     testDarknetLayer("shortcut_leaky");
+    testDarknetLayer("shortcut_unequal");
+    testDarknetLayer("shortcut_unequal_2");
 }
 
 TEST_P(Test_Darknet_layers, upsample)

modules/dnn/test/test_layers.cpp

@@ -1506,7 +1506,7 @@ TEST_P(Layer_Test_Eltwise_unequal, Accuracy)
     const int inpShapes[][4] = {{1, 4, 2, 2}, {1, 5, 2, 2}, {1, 3, 2, 2}};
     std::vector<String> inpNames(3);
     std::vector<Mat> inputs(3);
-    size_t numValues = 0;
+    size_t numOutValues = 1*4*2*2;  // By the first input
 
     std::vector<float> weights(3, 1);
     if (weighted)
@@ -1520,18 +1520,20 @@ TEST_P(Layer_Test_Eltwise_unequal, Accuracy)
     for (int i = 0; i < inputs.size(); ++i)
     {
         inputs[i].create(4, inpShapes[i], CV_32F);
-        numValues = std::max(numValues, inputs[i].total());
         randu(inputs[i], 0, 255);
         inpNames[i] = format("input_%d", i);
         net.connect(0, i, eltwiseId, i);
     }
-    Mat ref(1, numValues, CV_32F, Scalar(0));
+    Mat ref(1, numOutValues, CV_32F, Scalar(0));
 
     net.setInputsNames(inpNames);
     for (int i = 0; i < inputs.size(); ++i)
     {
         net.setInput(inputs[i], inpNames[i]);
-        ref.colRange(0, inputs[i].total()) += weights[i] * inputs[i].reshape(1, 1);
+        if (numOutValues >= inputs[i].total())
+            ref.colRange(0, inputs[i].total()) += weights[i] * inputs[i].reshape(1, 1);
+        else
+            ref += weights[i] * inputs[i].reshape(1, 1).colRange(0, numOutValues);
     }
 
     net.setPreferableBackend(backendId);