[ONNX] Unsqueeze operator (#1521)

src/ngraph/frontend/onnx_import/CMakeLists.txt

@@ -62,6 +62,8 @@ add_library(onnx_import STATIC
         op/split.hpp
         op/sub.hpp
         op/sum.hpp
+        op/unsqueeze.cpp
+        op/unsqueeze.hpp
         ops_bridge.cpp
         utils/broadcasting.cpp
         utils/broadcasting.hpp

src/ngraph/frontend/onnx_import/op/unsqueeze.cpp

+//*****************************************************************************
+// Copyright 2018 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#include <numeric>
+
+#include "unsqueeze.hpp"
+
+#include "ngraph/op/reshape.hpp"
+
+#include "exceptions.hpp"
+
+namespace ngraph
+{
+    namespace onnx_import
+    {
+        namespace op
+        {
+            NodeVector unsqueeze(const Node& node)
+            {
+                NodeVector inputs{node.get_ng_inputs()};
+                auto data = inputs.at(0);
+                auto data_shape = data->get_shape();
+                auto axes = node.get_attribute_value<std::vector<int64_t>>("axes");
+                if (axes.empty())
+                {
+                    throw error::parameter::Value(
+                        "Unsqueeze", node.get_name(), "axes attribute is mandatory.");
+                }
+                std::sort(std::begin(axes), std::end(axes), std::greater<int64_t>());
+
+                // Generate an increasing sequence (0,1,2,3..) as input_order for Reshape
+                AxisVector input_order(data_shape.size());
+                std::iota(std::begin(input_order), std::end(input_order), 0);
+
+                for (auto axis : axes)
+                {
+                    if ((axis < 0) || (axis > data_shape.size()))
+                    {
+                        throw error::parameter::Value(
+                            "Unsqueeze", node.get_name(), "provided axes attribute is not valid.");
+                    }
+                    data_shape.insert(std::next(std::begin(data_shape), axis), 1);
+                }
+
+                return {std::make_shared<ngraph::op::Reshape>(data, input_order, data_shape)};
+            }
+
+        } // namespace  op
+
+    } // namespace  onnx_import
+
+} // namespace  ngraph

src/ngraph/frontend/onnx_import/op/unsqueeze.hpp

+//*****************************************************************************
+// Copyright 2018 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#pragma once
+
+#include "ngraph/node_vector.hpp"
+
+#include "core/node.hpp"
+
+namespace ngraph
+{
+    namespace onnx_import
+    {
+        namespace op
+        {
+            NodeVector unsqueeze(const Node& node);
+        } // namespace  op
+
+    } // namespace  onnx_import
+
+} // namespace  ngraph

src/ngraph/frontend/onnx_import/ops_bridge.cpp

@@ -38,6 +38,7 @@
 #include "op/split.hpp"
 #include "op/sub.hpp"
 #include "op/sum.hpp"
+#include "op/unsqueeze.hpp"
 #include "ops_bridge.hpp"
 
 namespace ngraph
@@ -104,6 +105,7 @@ namespace ngraph
                     m_map.emplace("Split", std::bind(op::split, std::placeholders::_1));
                     m_map.emplace("Sub", std::bind(op::sub, std::placeholders::_1));
                     m_map.emplace("Sum", std::bind(op::sum, std::placeholders::_1));
+                    m_map.emplace("Unsqueeze", std::bind(op::unsqueeze, std::placeholders::_1));
                 }
 
                 NodeVector operator()(const Node& node) const

test/onnx_import.cpp

@@ -536,6 +536,29 @@ TEST(onnx, model_sub)
     EXPECT_TRUE(test::all_close_f(expected_output, result_vectors.front()));
 }
 
+TEST(onnx, model_unsqueeze)
+{
+    auto function = onnx_import::import_onnx_function(
+        file_util::path_join(SERIALIZED_ZOO, "onnx/unsqueeze.onnx"));
+
+    Inputs inputs;
+    inputs.emplace_back(test::NDArray<float, 3>(
+                            {{{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}},
+                             {{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}},
+                             {{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}}})
+                            .get_vector());
+
+    Outputs expected_output{
+        test::NDArray<float, 4>(
+            {{{{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}},
+              {{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}},
+              {{1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}, {1, 1, 1, 1, 1}}}})
+            .get_vector()};
+
+    Outputs outputs{execute(function, inputs, "INTERPRETER")};
+    EXPECT_TRUE(test::all_close_f(expected_output.front(), outputs.front()));
+}
+
 TEST(onnx, model_div)
 {
     auto function = ngraph::onnx_import::import_onnx_function(