diff --git a/modules/dnn/src/layers/einsum_layer.cpp b/modules/dnn/src/layers/einsum_layer.cpp
index 7f8eb47112..4e3eca2378 100644
--- a/modules/dnn/src/layers/einsum_layer.cpp
+++ b/modules/dnn/src/layers/einsum_layer.cpp
@@ -5,6 +5,7 @@
 #include <inttypes.h>
 #include <opencv2/dnn/shape_utils.hpp>
 #include "../precomp.hpp"
+#include "../ie_ngraph.hpp"
 #include "layers_common.hpp"
 #include "cpu_kernels/fast_gemm.hpp"
 
@@ -304,7 +305,7 @@ public:
     MatShape einsumOutDims; // vector to store output dimentions
 
     // These hold equation subring, left hand side and right it of
-    String lhs_eq, rhs_eq;
+    String lhs_eq, rhs_eq, equation;
 
     // Holds token from left hand side of the equation
     std::vector<String> lhs_eq_tokens;
@@ -378,7 +379,7 @@ public:
     LayerEinsumImpl(const LayerParams& params)
     {
         setParamsFrom(params);
-        String equation = params.get<String>("equation");
+        equation = params.get<String>("equation");
         int outputSize = params.get<int>("outputSize");
         numInputs  = params.get<int>("inputSize");
 
@@ -423,6 +424,11 @@ public:
         calculateOutputShape();
     }
 
+    virtual bool supportBackend(int backendId) CV_OVERRIDE {
+        return backendId == DNN_BACKEND_OPENCV ||
+               backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH;
+    }
+
     // getMeoryShapes
     bool getMemoryShapes(const std::vector<MatShape> &inputs,
                          const int requiredOutputs,
@@ -553,6 +559,19 @@ public:
         result = result.reshape(1, einsumOutDims.size(), einsumOutDims.data());
         result.copyTo(outputs[0]);
     } // forward
+
+#ifdef HAVE_DNN_NGRAPH
+    virtual Ptr<BackendNode> initNgraph(const std::vector<Ptr<BackendWrapper> >&,
+                                        const std::vector<Ptr<BackendNode> >& nodes) CV_OVERRIDE {
+        ov::OutputVector inputs(nodes.size());
+        for (size_t i = 0; i < nodes.size(); ++i) {
+            inputs[i] = nodes[i].dynamicCast<InfEngineNgraphNode>()->node;
+        }
+        auto einsum = std::make_shared<ov::op::v7::Einsum>(inputs, equation);
+        return new InfEngineNgraphNode(einsum);
+    }
+#endif // HAVE_DNN_NGRAPH
+
 }; // EinsumClass
 
 Mat LayerEinsumImpl::reduceSum(Mat& src, MatShape& reduceAxis)
diff --git a/modules/dnn/test/test_onnx_conformance_layer_filter__openvino.inl.hpp b/modules/dnn/test/test_onnx_conformance_layer_filter__openvino.inl.hpp
index 291ea30e92..509cf6007d 100644
--- a/modules/dnn/test/test_onnx_conformance_layer_filter__openvino.inl.hpp
+++ b/modules/dnn/test/test_onnx_conformance_layer_filter__openvino.inl.hpp
@@ -597,7 +597,7 @@ CASE(test_dynamicquantizelinear_min_adjusted_expanded)
 CASE(test_edge_pad)
     // no filter
 CASE(test_einsum_batch_diagonal)
-    // no filter
+    SKIP;
 CASE(test_einsum_batch_matmul)
     // no filter
 CASE(test_einsum_inner_prod)
diff --git a/modules/dnn/test/test_onnx_importer.cpp b/modules/dnn/test/test_onnx_importer.cpp
index 859757d17a..4d475857e5 100644
--- a/modules/dnn/test/test_onnx_importer.cpp
+++ b/modules/dnn/test/test_onnx_importer.cpp
@@ -1471,6 +1471,8 @@ TEST_P(Test_ONNX_layers, Einsum_2D)
 
 TEST_P(Test_ONNX_layers, Einsum_2D_Ellipses)
 {
+    if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
+        applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
     testONNXModels("einsum_2d_ellipses", npy, 0, 0, false, false, 2);
 }
 
@@ -1501,6 +1503,8 @@ TEST_P(Test_ONNX_layers, DISABLED_Einsum_HadamardProduct)
 
 TEST_P(Test_ONNX_layers, Einsum_Batch_Diagonal)
 {
+    if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
+        applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
     testONNXModels("einsum_batch_diagonal", npy, 0, 0, false, false, 1);
 }