diff --git a/src/frontends/tensorflow/src/input_model.cpp b/src/frontends/tensorflow/src/input_model.cpp
index cd64c466dec1b3..14474eec64125a 100644
--- a/src/frontends/tensorflow/src/input_model.cpp
+++ b/src/frontends/tensorflow/src/input_model.cpp
@@ -376,11 +376,12 @@ std::vector<std::shared_ptr<OpPlace>> InputModel::InputModelTFImpl::topologicall
         ops_to_do.push(output_operation_place);
     }
 
-    // walk through all NextIteration nodes and put their producers into ops_to_do
-    // this is needed to avoid missed nodes in the body graph of TF1 While operation
     for (const auto& op_place : m_op_places) {
         auto op_decoder = op_place->get_decoder();
+        auto op_name = op_decoder->get_op_name();
         if (op_decoder->get_op_type() == "NextIteration") {
+            // walk through all NextIteration nodes and put their producers into ops_to_do
+            // this is needed to avoid missed nodes in the body graph of TF1 While operation
             std::string producer_name;
             std::string producer_output_port_name;
             size_t producer_output_port_idx;
@@ -390,6 +391,15 @@ std::vector<std::shared_ptr<OpPlace>> InputModel::InputModelTFImpl::topologicall
                                     "NextIteration is not found among operation places " +
                                         producer_name);
             ops_to_do.push(m_op_places_map.at(producer_name));
+        } else if (op_decoder->get_op_type() == "LookupTableImport" ||
+                   op_decoder->get_op_type() == "LookupTableImportV2") {
+            // all LookupTableImport nodes must be preserved in a graph for conversion because
+            // they can be terminating nodes and contain input values for HashTable initialization
+            FRONT_END_GENERAL_CHECK(m_op_places_map.count(op_name),
+                                    "[TensorFlow Frontend] internal error or inconsistent model: LookupTableImport "
+                                    "operation is not found among operation places " +
+                                        op_name);
+            ops_to_do.push(m_op_places_map.at(op_name));
         }
     }
 
diff --git a/tests/model_hub_tests/models_hub_common/utils.py b/tests/model_hub_tests/models_hub_common/utils.py
index 9c5a02207b5729..d223f3bc984c6e 100644
--- a/tests/model_hub_tests/models_hub_common/utils.py
+++ b/tests/model_hub_tests/models_hub_common/utils.py
@@ -56,11 +56,26 @@ def get_models_list_not_skipped(model_list_file: str, skip_list_file: str):
 
 def compare_two_tensors(ov_res, fw_res, eps):
     is_ok = True
-    if not np.allclose(ov_res, fw_res, atol=eps, rtol=eps, equal_nan=True):
+    if ov_res.dtype.type == str or ov_res.dtype.type == np.str_ or ov_res.dtype.type == np.object_:
+        ov_res = ov_res.astype('U')
+        # TF can represent string tensors in different format: array of bytestreams
+        # so we have to align formats of both string tensors, for example, to unicode
+        if ov_res.dtype.type != fw_res.dtype.type:
+            try:
+                fw_res = fw_res.astype('U')
+            except:
+                # ref_array of object type and each element must be utf-8 decoded
+                utf8_decoded_elems = [elem.decode('UTF-8') for elem in fw_res.flatten()]
+                fw_res = np.array(utf8_decoded_elems, dtype=str).reshape(fw_res.shape)
+        is_ok = np.array_equal(ov_res, fw_res)
+    elif ov_res.dtype == bool:
+        is_ok = np.array_equal(ov_res, fw_res)
+    elif not np.allclose(ov_res, fw_res, atol=eps, rtol=eps, equal_nan=True):
         is_ok = False
         max_diff = np.abs(ov_res.astype(np.float32) - fw_res.astype(np.float32)).max()
         print("Max diff is {}".format(max_diff))
-    else:
+
+    if is_ok:
         print("Accuracy validation successful!\n")
         print("absolute eps: {}, relative eps: {}".format(eps, eps))
     return is_ok
diff --git a/tests/model_hub_tests/tensorflow/model_lists/precommit_read_model b/tests/model_hub_tests/tensorflow/model_lists/precommit_read_model
index 8e24ef1cbd8833..2a40914e660d84 100644
--- a/tests/model_hub_tests/tensorflow/model_lists/precommit_read_model
+++ b/tests/model_hub_tests/tensorflow/model_lists/precommit_read_model
@@ -4,4 +4,6 @@ mil-nce/s3d,https://www.kaggle.com/models/deepmind/mil-nce/frameworks/tensorFlow
 yamnet,https://www.kaggle.com/models/google/yamnet/frameworks/tensorFlow2/variations/yamnet/versions/1
 universal-sentence-encoder-multilingual,https://www.kaggle.com/models/google/universal-sentence-encoder/frameworks/tensorFlow2/variations/multilingual/versions/2
 movenet/singlepose/lightning,https://www.kaggle.com/models/google/movenet/frameworks/tensorFlow2/variations/singlepose-lightning/versions/4
-imagenet/resnet_v2_50/feature_vector,https://www.kaggle.com/models/google/resnet-v2/frameworks/tensorFlow2/variations/50-feature-vector/versions/2
\ No newline at end of file
+imagenet/resnet_v2_50/feature_vector,https://www.kaggle.com/models/google/resnet-v2/frameworks/tensorFlow2/variations/50-feature-vector/versions/2
+# LookupTableImportV2 is terminating node and is needed for conversion
+openimages_v4/ssd/mobilenet_v2,https://www.kaggle.com/models/google/mobilenet-v2/frameworks/tensorFlow1/variations/openimages-v4-ssd-mobilenet-v2/versions/1
\ No newline at end of file