diff --git a/research/cv/delf/README_CN.md b/research/cv/delf/README_CN.md
index 956eac9a6d83178d9495fd74820dd70c7a392457..df3dd6716980232e1d2301eeb06b14349a9cddc1 100755
--- a/research/cv/delf/README_CN.md
+++ b/research/cv/delf/README_CN.md
@@ -86,6 +86,9 @@
 - 如需查看详情，请参见如下资源：
     - [MindSpore教程](https://www.mindspore.cn/tutorials/zh-CN/master/index.html)
     - [MindSpore Python API](https://www.mindspore.cn/docs/api/zh-CN/master/index.html)
+- 依赖包
+    - h5py版本：3.4.0
+    - 其它依赖包信息详见requirements.txt文件
 
 # 快速入门
 
@@ -93,7 +96,7 @@
 
 - 数据集下载和预处理
 
-  ```yaml
+  ```shell
   # Google Landmarks Dataset v2 训练集下载以及转化为mindrecord文件
   # 【注】请准备至少1.1TB的存储空间，若空间不足可以将可选参数[NEED_ROMOVE_TAR]设置为'y'，设置后占用约633G存储空间
   bash scripts/download_gldv2.sh 500 [DATASET_PATH] [NEED_ROMOVE_TAR]
@@ -104,9 +107,9 @@
   bash scripts/download_paris.sh [DATASET_PATH]
   ```
 
-- 迁移学习预训练权重下载
+- 预训练权重下载
 
-  ```yaml
+  ```shell
   # 下载ImageNet预训练的Resnet50权重和pca降维预训练转换矩阵
   bash scripts/download_pretrained.sh
   ```
@@ -121,12 +124,14 @@
   ```
 
   ```shell
+  # 请确保前面步骤的预训练权重已经下载完毕！
   # 运行训练示例，分为两阶段进行训练
   # 微调阶段：
-  python train.py --train_state=tuning > train_tuning.log 2>&1 &
+  bash scripts/run_1p_train.sh tuning
   # 注意力训练阶段：
-  # 修改checkpoint_path为微调阶段得到的checkpoint
-  python train.py --train_state=attn --checkpoint_path=/home/delf/ckpt/checkpoint_delf_tuning-1_4989.ckpt > train_attn.log 2>&1 &
+  # 修改[CHECKPOINT]为微调阶段得到的checkpoint文件路径
+  bash scripts/run_1p_train.sh attn [CHECKPOINT]
+  # example: bash scripts/run_1p_train.sh attn ./ckpt/checkpoint_delf_tuning-1_4989.ckpt
 
   # 运行分布式训练示例
   # 微调阶段
@@ -208,6 +213,7 @@
     │   ├── download_oxf.sh                                  # 下载 Oxford5k 的shell脚本
     │   ├── download_paris.sh                                # 下载 Paris6k 的shell脚本
     │   ├── download_pretrained.sh                           # 下载预训练权重的shell脚本
+    │   ├── run_1p_train.sh                                  # 单卡训练的shell脚本
     │   ├── run_8p_train.sh                                  # 分布式训练的shell脚本
     │   ├── run_eval_match_images.sh                         # 图像匹配评估的shell脚本
     │   ├── run_eval_retrieval_images.sh                     # 图像检索评估的shell脚本
@@ -350,18 +356,19 @@ bash scripts/download_paris.sh [DATASET_PATH]
   为了让模型达到更好更快地收敛，delf模型需要进行两个阶段的训练，微调阶段和注意力训练阶段，微调阶段运行命令：
 
   ```bash
-  python train.py --train_state=tuning > train_tuning.log 2>&1 &
+  bash scripts/run_1p_train.sh tuning
   ```
 
   注意力训练阶段运行命令：
 
   ```bash
-  python train.py --train_state=attn --checkpoint_path=/home/delf/ckpt/checkpoint_delf_tuning-1_4989.ckpt > train_attn.log 2>&1 &
+  bash scripts/run_1p_train.sh attn [CHECKPOINT]
+  # example: bash scripts/run_1p_train.sh attn ./ckpt/checkpoint_delf_tuning-1_4989.ckpt
   ```
 
   需要注意的是注意力训练阶段要载入微调阶段已经训练好的checkpoint权重。
 
-  上述python命令将在后台运行，您可以通过train_tuning.log或train_attn.log文件查看结果。
+  上述命令将在后台运行，您可以通过train_tuning.log或train_attn.log文件查看结果。
 
   训练结束后，您可在默认脚本文件夹的`ckpt`目录下找到检查点文件。
 
@@ -424,7 +431,7 @@ bash scripts/download_paris.sh [DATASET_PATH]
   bash scripts/run_eval_match_images.sh [IMAGES_PATH] [CHECKPOINT] [DEVICES]
   ```
 
-  上述python命令将在后台打印日志，您可以通过`extract_feature.log`文件查看特征提取情况，通过`match_images.log`文件查看特征匹配情况。最后得到展示两个图像匹配特征的图片，它会保存为默认脚本目录下的`eval_match.png`。
+  上述命令将在后台打印日志，您可以通过`extract_feature.log`文件查看特征提取情况，通过`match_images.log`文件查看特征匹配情况。最后得到展示两个图像匹配特征的图片，它会保存为默认脚本目录下的`eval_match.png`。
 
   注：对于分布式训练后评估，请将checkpoint_path设置为最后保存的检查点文件，如`/username/delf/train_parallel0/ckpt/attn-1-4898.ckpt`。
 
@@ -440,7 +447,7 @@ bash scripts/download_paris.sh [DATASET_PATH]
   bash scripts/run_eval_retrieval_images.sh [IMAGES_PATH] [GT_PATH] [CHECKPOINT] [DEVICES]
   ```
 
-  上述python命令将在后台打印日志，您可以通过`extract_feature.log`文件查看特征提取情况，通过`./retrieval_dataset/process[X]/retrieval[X].log`文件查看检索情况，通过`calculate_mAP.log`文件查看计算mAP的情况。计算得到的mAP结果会在脚本运行完成后打印在终端，也会保存到默认脚本目录下的mAP.txt文件中：
+  上述命令将在后台打印日志，您可以通过`extract_feature.log`文件查看特征提取情况，通过`./retrieval_dataset/process[X]/retrieval[X].log`文件查看检索情况，通过`calculate_mAP.log`文件查看计算mAP的情况。计算得到的mAP结果会在脚本运行完成后打印在终端，也会保存到默认脚本目录下的mAP.txt文件中：
 
   ```matlab
   # cat mAP.txt
diff --git a/research/cv/delf/delf_config.yaml b/research/cv/delf/delf_config.yaml
index a6588731324cb68f1b473318605fe1e9b35121e6..3066b9316a44fee583682d3160c25e40c619bdf5 100755
--- a/research/cv/delf/delf_config.yaml
+++ b/research/cv/delf/delf_config.yaml
@@ -25,7 +25,7 @@ attention_loss_weight: 1.0
 
 traindata_path: "/mass_data/gldv2_dataset/mindrecord/train.mindrecord000"
 keep_checkpoint_max: 1
-imagenet_checkpoint: "/home/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5"
+imagenet_checkpoint: "./resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5"
 checkpoint_path: ""
 save_ckpt: "./ckpt/"
 save_ckpt_step: 2000
diff --git a/research/cv/delf/requirements.txt b/research/cv/delf/requirements.txt
index cba88de4dea5048690a5f32dd05eef9e3dada21c..cd8ea1509c47a2e8073ef9557d57bf6d3bfef4d2 100755
--- a/research/cv/delf/requirements.txt
+++ b/research/cv/delf/requirements.txt
@@ -3,7 +3,7 @@ annoy
 pillow
 csv
 pandas
-h5py
+h5py==3.4.0
 matplotlib
 scipy
 scikit-image
\ No newline at end of file
diff --git a/research/cv/delf/scripts/run_1p_train.sh b/research/cv/delf/scripts/run_1p_train.sh
new file mode 100644
index 0000000000000000000000000000000000000000..0a32cebeb7f706eaf482430e4ed42016351dcf31
--- /dev/null
+++ b/research/cv/delf/scripts/run_1p_train.sh
@@ -0,0 +1,51 @@
+#!/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 [[ $# -lt 1 || $# -gt 2 ]]
+then
+    echo "Usage: bash scripts/run_1p_train.sh [TRAIN_STATE] [CHECKPOINT]"
+exit 1
+fi
+
+if [ "$1" == "tuning" ] || [ "$1" == "attn" ];then
+    train_state=$1
+else
+    echo "train state, it's value must be in [tuning, attn]"
+    exit 1
+fi
+
+checkpoint=""
+if [ $# == 2 ]
+then
+    checkpoint=$2
+fi
+
+cores=`cat /proc/cpuinfo|grep "processor" |wc -l`
+echo "the number of logical core" $cores
+
+EXECUTE_PATH=$(pwd)
+config_path="${EXECUTE_PATH}/delf_config.yaml"
+
+export DEVICE_ID=0
+export RANK_ID=0
+export RANK_SIZE=1
+
+echo "Start training for rank 0, device 0"
+
+python train.py \
+--config_path=$config_path \
+--train_state=$train_state \
+--checkpoint_path=$checkpoint > $train_state.log 2>&1 &
diff --git a/research/cv/delf/src/convert_h5_to_weight.py b/research/cv/delf/src/convert_h5_to_weight.py
index f69b1a13443c20b861f3bf7a1f13c0828ea6a1fc..8b1a52b44f87615eba361bd392478ae6b0ebfbd5 100755
--- a/research/cv/delf/src/convert_h5_to_weight.py
+++ b/research/cv/delf/src/convert_h5_to_weight.py
@@ -33,7 +33,7 @@ def translate_h5(h5path=''):
             param_name = param_name.replace('_beta_1:0', '.beta')
             param_name = param_name.replace('_running_mean_1:0', '.moving_mean')
             param_name = param_name.replace('_running_std_1:0', '.moving_variance')
-            data = f[i][name].value.astype(np.float32)
+            data = f[i][name][:].astype(np.float32)
             count += 1
             weights[param_name] = data
     parameter_dict = {}