diff --git a/research/cv/UNet3+/README_CN.md b/research/cv/UNet3+/README_CN.md
index 02c74fe36e6c0cfd6f129a7dceef70738adffd4b..8fa0d505389d66a0d54a630099bb2339dc6ded66 100644
--- a/research/cv/UNet3+/README_CN.md
+++ b/research/cv/UNet3+/README_CN.md
@@ -18,6 +18,7 @@
     - [Mindir推理](#Mindir推理)
         - [导出模型](#导出模型)
         - [在Ascend310执行推理](#在Ascend310执行推理)
+        - [onnx推理](#onnx推理)
         - [结果](#结果)
 - [模型描述](#模型描述)
     - [性能](#性能)
@@ -107,6 +108,7 @@ Ascend训练：生成[RANK_TABLE_FILE](https://gitee.com/mindspore/models/tree/m
         │   ├──run_eval.sh                   // 推理启动脚本
         │   ├──run_train.sh                  // 训练启动脚本
         |   ├──run_infer_310.sh              // 310推理启动脚本
+            |——run_eval_onnx.sh              // onnx推理启动脚本
         ├── src
         │   ├──config.py                     // 配置加载文件
         │   ├──dataset.py                    // 数据集处理
@@ -120,6 +122,7 @@ Ascend训练：生成[RANK_TABLE_FILE](https://gitee.com/mindspore/models/tree/m
         ├── dataset_preprocess.py            // 数据集预处理脚本
         ├── postprocess.py                   // 310精度计算脚本
         ├── preprocess.py                    // 310预处理脚本
+        |—— eval_onnx.py                     // onnx推理脚本
 
 ```
 
@@ -234,11 +237,11 @@ Ascend训练：生成[RANK_TABLE_FILE](https://gitee.com/mindspore/models/tree/m
 ### [导出MindIR](#contents)
 
 ```shell
-python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
+python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT] --device_target [DEVICE_TARGET]
 ```
 
 - `ckpt_file`为必填项。
-- `file_format` 必须在 ["AIR", "MINDIR"]中选择。
+- `file_format` 必须在 ["AIR", "MINDIR","ONNX"]中选择。
 
 ### 在Ascend310执行推理
 
@@ -253,6 +256,19 @@ bash run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [NEED_PREPROCESS] [DEVICE_ID]
 - `NEED_PREPROCESS` 表示数据是否需要预处理，取值范围为 'y' 或者 'n'。
 - `DEVICE_ID` 可选，默认值为0。
 
+### onnx推理
+
+在执行推理前，onnx文件必须通过`export.py`脚本导出。
+
+```shell
+# onnx inference
+bash run_eval_onnx.sh [DATASET_PATH] [ONNX_MODEL] [DEVICE_TARGET]
+```
+
+- `DATASET_PATH` 为测试集所在的路径
+- `ONNX_MODEL` 为导出的onnx模型所在路径
+- `DEVICE_TARGET` 必须在['Ascend','CPU','GPU']中选择
+
 ### 结果
 
 推理结果保存在脚本执行的当前路径，你可以在acc.log中看到以下精度计算结果。
diff --git a/research/cv/UNet3+/eval_onnx.py b/research/cv/UNet3+/eval_onnx.py
new file mode 100644
index 0000000000000000000000000000000000000000..1cc7a018bc29152c4384b9a142a7f27de7466161
--- /dev/null
+++ b/research/cv/UNet3+/eval_onnx.py
@@ -0,0 +1,76 @@
+# Copyright 2022 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+"""Run evaluation for a model exported to ONNX"""
+import os
+import datetime
+import onnxruntime as ort
+import numpy as np
+from mindspore.common import set_seed
+
+from src.logger import get_logger
+from src.dataset import create_Dataset
+from src.config import config
+from eval import AverageMeter, dice_coef
+
+
+def sigmoid(prediction):
+    """sigmoid function"""
+    negative = prediction < 0
+    positive = prediction > 0
+    sig_prediction = np.zeros_like(prediction)
+    sig_prediction[negative] = np.exp(prediction[negative]) / (1 + np.exp(prediction[negative]))
+    sig_prediction[positive] = 1 / (1 + np.exp(-prediction[positive]))
+    return sig_prediction
+
+
+def create_session(checkpoint_path, target_device):
+    """create_session for onnx"""
+    if target_device == "GPU":
+        providers = ['CUDAExecutionProvider']
+    elif target_device in ['CPU', 'Ascend']:
+        providers = ['CPUExecutionProvider']
+    else:
+        raise ValueError(f"Unsupported target device '{target_device}'.\
+         Expected one of: 'CPU', 'GPU', 'Ascend'")
+    session = ort.InferenceSession(checkpoint_path, providers=providers)
+    input_name = session.get_inputs()[0].name
+    return session, input_name
+
+
+def run_onnx_eval():
+    """run_onnx_eval"""
+    set_seed(1)
+    config.save_dir = os.path.join(config.output_path,
+                                   datetime.datetime.now().strftime('%Y-%m-%d_time_%H_%M_%S'))
+    if not os.path.exists(config.save_dir):
+        os.makedirs(config.save_dir)
+    config.logger = get_logger(config.save_dir, "UNet3Plus", 0)
+    config.logger.save_args(config)
+    dataset, _ = create_Dataset(data_path=config.val_data_path, aug=0,
+                                batch_size=config.batch_size,
+                                device_num=1, rank=0, shuffle=False)
+    data_loader = dataset.create_dict_iterator(num_epochs=1, output_numpy=True)
+    dices = AverageMeter()
+    session, input_name = create_session(config.file_name, config.device_target)
+    for episode, data in enumerate(data_loader):
+        y_pred = sigmoid(session.run(None, {input_name: data['image']})[0])
+        dice = dice_coef(y_pred, data['mask'])
+        config.logger.info("episode %d : dice = %f ", episode, dice)
+        dices.update(dice, config.batch_size)
+    config.logger.info("Final dice: %s", str(dices.avg))
+
+
+if __name__ == "__main__":
+    run_onnx_eval()
diff --git a/research/cv/UNet3+/scripts/run_eval_onnx.sh b/research/cv/UNet3+/scripts/run_eval_onnx.sh
new file mode 100644
index 0000000000000000000000000000000000000000..bef02703f9e1f25d20dafec0bbe92f31c9187831
--- /dev/null
+++ b/research/cv/UNet3+/scripts/run_eval_onnx.sh
@@ -0,0 +1,47 @@
+#!/bin/bash
+# Copyright 2022 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+if [ $# != 3 ];then
+  echo "Usage:bash scripts/eval_onnx.sh [DATASET_PATH] [ONNX_MODEL] [DEVICE_TARGET]"
+  exit 1
+fi
+
+
+get_real_path(){
+  if [ "${1:0:1}" == "/" ];then
+    echo "$1"
+  else
+    realpath -m "$PWD/$1"
+  fi
+}
+
+DATASET_PATH=$(get_real_path "$1")
+ONNX_MODEL=$(get_real_path "$2")
+DEVICE_TARGET=${3}
+
+if [ ! -d "$DATASET_PATH" ];then
+  echo "error:DATASET_PATH=${DATASET_PATH} is not a directory"
+  exit 1
+fi
+
+if [ ! -f "$ONNX_MODEL" ];then
+  echo "error:ONNX_MODEL=${ONNX_MODEL} is not a file"
+  exit 1
+fi
+
+python ../eval_onnx.py \
+  --val_data_path="$DATASET_PATH" \
+  --file_name="$ONNX_MODEL" \
+  --device_target="$DEVICE_TARGET" > eval_onnx.log 2>&1 &
diff --git a/research/cv/UNet3+/src/dataset.py b/research/cv/UNet3+/src/dataset.py
index eb8ccb3df182e78797906b8c19c294424d509602..d2b2d75ca201e052879f80959574ffdddfb04b2c 100644
--- a/research/cv/UNet3+/src/dataset.py
+++ b/research/cv/UNet3+/src/dataset.py
@@ -18,7 +18,6 @@ import glob
 import random
 import numpy as np
 from skimage.io import imread
-from skimage import color
 
 import mindspore.dataset as ds
 import mindspore.dataset.vision as CV
@@ -48,7 +47,6 @@ class Dataset:
                 image = image[::-1, :, :].copy()
                 mask = mask[::-1, :].copy()
 
-        image = color.gray2rgb(image)
         mask = mask[:, :, np.newaxis]
         return image, mask