diff --git a/research/cv/ProtoNet/README.md b/research/cv/ProtoNet/README.md
index 84d821c65bfb810493c041f428bbe3e7f00fe777..4f17b2f33bddf91e981771412879e7b7fe094ddd 100644
--- a/research/cv/ProtoNet/README.md
+++ b/research/cv/ProtoNet/README.md
@@ -157,6 +157,38 @@ Test Acc in Ascend: 0.9954400658607483  Loss: 0.02102319709956646
 Test Acc in GPU: 0.996999979019165  Loss: 0.013885765336453915
 ```
 
+## [Inference Process](#contents)
+
+### [Export MindIR](#contents)
+
+```shell
+python export.py --ckpt_file [CKPT_PATH] --file_format [FILE_FORMAT]
+```
+
+The ckpt_file parameter is required,
+`EXPORT_FORMAT` should be in ["AIR", "MINDIR"]
+
+### [Infer on Ascend310](#contents)
+
+Before performing inference, the mindir file must be exported by `export.py` script. We only provide an example of inference using MINDIR model.
+
+```shell
+# Ascend310 inference
+bash run_infer_310.sh [MINDIR_PATH] [DATASET_PATH] [DEVICE_ID]
+```
+
+- `MINDIR_PATH` specifies path of used "MINDIR" OR "AIR" model.
+- `DATASET_PATH` specifies path of omniglot datasets  
+- `DEVICE_ID` is optional, default value is 0.
+
+### [Result](#contents)
+
+Inference result is saved in current path, you can find result like this in acc.log file.
+
+```bash
+'acc': 0.9956
+```
+
 # [Model Description](#contents)
 
 ## [Performance](#contents)
diff --git a/research/cv/ProtoNet/ascend310_infer/CMakeLists.txt b/research/cv/ProtoNet/ascend310_infer/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d580e7e6d96afc8ff0449eef913f908585e46e67
--- /dev/null
+++ b/research/cv/ProtoNet/ascend310_infer/CMakeLists.txt
@@ -0,0 +1,20 @@
+cmake_minimum_required(VERSION 3.14.1)
+project(Ascend310Infer)
+add_compile_definitions(_GLIBCXX_USE_CXX11_ABI=0)
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O0 -g -std=c++17 -Werror -Wall -fPIE -Wl,--allow-shlib-undefined")
+
+set(PROJECT_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR}/)
+
+option(MINDSPORE_PATH "mindspore install path" "")
+include_directories(${MINDSPORE_PATH})
+include_directories(${MINDSPORE_PATH}/include)
+
+include_directories(${PROJECT_SRC_ROOT})
+
+find_library(MS_LIB libmindspore.so ${MINDSPORE_PATH}/lib)
+file(GLOB_RECURSE MD_LIB ${MINDSPORE_PATH}/_c_dataengine*)
+find_package(gflags REQUIRED)
+
+add_executable(main src/main.cc src/utils.cc)
+target_link_libraries(main ${MS_LIB} ${MD_LIB} gflags)
+
diff --git a/research/cv/ProtoNet/ascend310_infer/build.sh b/research/cv/ProtoNet/ascend310_infer/build.sh
new file mode 100644
index 0000000000000000000000000000000000000000..770a8851efade7f352039fc8665d307ae1abbb00
--- /dev/null
+++ b/research/cv/ProtoNet/ascend310_infer/build.sh
@@ -0,0 +1,23 @@
+#!/bin/bash
+# Copyright 2021 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 [ ! -d out ]; then
+  mkdir out
+fi
+cd out || exit
+cmake .. \
+    -DMINDSPORE_PATH="`pip show mindspore-ascend | grep Location | awk '{print $2"/mindspore"}' | xargs realpath`"
+make
diff --git a/research/cv/ProtoNet/ascend310_infer/inc/utils.h b/research/cv/ProtoNet/ascend310_infer/inc/utils.h
new file mode 100644
index 0000000000000000000000000000000000000000..52f9eb4354787c4cc67fc3c284bd849e29db5e86
--- /dev/null
+++ b/research/cv/ProtoNet/ascend310_infer/inc/utils.h
@@ -0,0 +1,41 @@
+/**
+ * Copyright 2021 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.
+ */
+
+#ifndef MINDSPORE_INFERENCE_UTILS_H_
+#define MINDSPORE_INFERENCE_UTILS_H_
+
+#include <sys/stat.h>
+#include <dirent.h>
+#include <vector>
+#include <string>
+#include <memory>
+#include "include/api/types.h"
+
+namespace ms = mindspore;
+// using namespace std;
+using std::vector;
+using std::string;
+using std::string_view;
+
+
+vector<string> GetAllFiles(string_view dir_name);
+DIR *OpenDir(string_view dir_name);
+string RealPath(string_view path);
+ms::MSTensor ReadFile(const string &file);
+size_t GetMax(ms::MSTensor data);
+int WriteResult(const string& imageFile, const vector<mindspore::MSTensor> &outputs);
+
+#endif
diff --git a/research/cv/ProtoNet/ascend310_infer/src/main.cc b/research/cv/ProtoNet/ascend310_infer/src/main.cc
new file mode 100644
index 0000000000000000000000000000000000000000..ac2b9e9038ce4508bde3b1f62c31be81280c99dd
--- /dev/null
+++ b/research/cv/ProtoNet/ascend310_infer/src/main.cc
@@ -0,0 +1,135 @@
+/**
+ * Copyright 2021 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.
+ */
+
+#include <sys/time.h>
+#include <gflags/gflags.h>
+#include <dirent.h>
+#include <iostream>
+#include <string>
+#include <algorithm>
+#include <iosfwd>
+#include <vector>
+#include <fstream>
+#include <sstream>
+
+#include "include/api/context.h"
+#include "include/api/model.h"
+#include "include/api/serialization.h"
+#include "include/dataset/execute.h"
+#include "include/api/types.h"
+#include "include/dataset/vision_ascend.h"
+#include "include/dataset/vision.h"
+#include "../inc/utils.h"
+
+namespace ms = mindspore;
+namespace ds = mindspore::dataset;
+
+
+DEFINE_string(mindir_path, ".", "mindir path");
+DEFINE_string(dataset_path, ".", "dataset path");
+DEFINE_int32(device_id, 0, "device id");
+
+
+
+
+int main(int argc, char **argv) {
+  using std::cout;
+  using std::endl;
+  using std::string;
+  using std::vector;
+  using std::make_shared;
+  using std::ofstream;
+  using std::stringstream;
+  using std::map;
+  using std::pair;
+  using std::ios;
+  gflags::ParseCommandLineFlags(&argc, &argv, true);
+  cout << FLAGS_mindir_path << endl;
+  cout << FLAGS_dataset_path << endl;
+  cout << FLAGS_device_id << endl;
+  // set context
+  auto context = make_shared<ms::Context>();
+  auto ascend310_info = make_shared<ms::Ascend310DeviceInfo>();
+  ascend310_info->SetDeviceID(FLAGS_device_id);
+  context->MutableDeviceInfo().push_back(ascend310_info);
+  // define model
+  ms::Graph graph;
+  ms::Status ret = ms::Serialization::Load(FLAGS_mindir_path, ms::ModelType::kMindIR, &graph);
+  if (ret != ms::kSuccess) {
+    cout << "Load model failed." << endl;
+    return 1;
+  }
+  ms::Model protoNet;
+  // build model
+  ret = protoNet.Build(ms::GraphCell(graph), context);
+  if (ret != ms::kSuccess) {
+    cout << "Build model failed." << endl;
+    return 1;
+  }
+  // get model input info
+  vector<ms::MSTensor> model_inputs = protoNet.GetInputs();
+  if (model_inputs.empty()) {
+    cout << "Invalid model, inputs is empty." << endl;
+    return 1;
+  }
+  // load data
+  vector<string> images = GetAllFiles(FLAGS_dataset_path);
+  map<double, double> costTime_map;
+  size_t size = images.size();
+  // start infer
+  for (size_t i = 0; i < size; ++i) {
+    struct timeval start = {0};
+    struct timeval end = {0};
+    double startTimeMs;
+    double endTimeMs;
+    vector<ms::MSTensor> outputs;
+    vector<ms::MSTensor> inputs;
+    auto image = ReadFile(images[i]);
+    inputs.emplace_back(model_inputs[0].Name(), model_inputs[0].DataType(), model_inputs[0].Shape(),
+                        image.Data().get(), image.DataSize());
+    gettimeofday(&start, nullptr);
+    ret = protoNet.Predict(inputs, &outputs);
+    gettimeofday(&end, nullptr);
+    if (ret != ms::kSuccess) {
+      cout << "Predict model failed." << endl;
+      return 1;
+    }
+    startTimeMs = (1.0 * start.tv_sec * 1000000 + start.tv_usec) / 1000;
+    endTimeMs = (1.0 * end.tv_sec * 1000000 + end.tv_usec) / 1000;
+    costTime_map.insert(pair<double, double>(startTimeMs, endTimeMs));
+    WriteResult(images[i], outputs);
+  }
+  double average = 0.0;
+  int inferCount = 0;
+  for (auto iter = costTime_map.begin(); iter != costTime_map.end(); iter++) {
+    double diff = 0.0;
+    diff = iter->second - iter->first;
+    average += diff;
+    inferCount++;
+  }
+  average = average / inferCount;
+  stringstream timeCost;
+  timeCost << "NN inference cost average time: "<< average << " ms of infer_count " << inferCount << endl;
+  cout << "NN inference cost average time: "<< average << "ms of infer_count " << inferCount << endl;
+  string fileName = "./time_Result" + string("/test_perform_static.txt");
+  ofstream fileStream(fileName.c_str(), ios::trunc);
+  fileStream << timeCost.str();
+  fileStream.close();
+  costTime_map.clear();
+  return 0;
+}
+
+
diff --git a/research/cv/ProtoNet/ascend310_infer/src/utils.cc b/research/cv/ProtoNet/ascend310_infer/src/utils.cc
new file mode 100644
index 0000000000000000000000000000000000000000..6469c6396420b0654577ba47c8aeadf960ba5189
--- /dev/null
+++ b/research/cv/ProtoNet/ascend310_infer/src/utils.cc
@@ -0,0 +1,170 @@
+/**
+ * Copyright 2021 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.
+ */
+
+#include <fstream>
+#include <algorithm>
+#include <iostream>
+#include "../inc/utils.h"
+
+namespace ms = mindspore;
+using mindspore::MSTensor;
+using std::vector;
+using std::string;
+using std::string_view;
+using std::sort;
+using std::shared_ptr;
+using std::cout;
+using std::endl;
+using std::ifstream;
+using std::ios;
+
+vector<string> GetAllFiles(string_view dir_name) {
+  struct dirent *filename;
+  DIR *dir = OpenDir(dir_name);
+  if (dir == nullptr) {
+    return {};
+  }
+
+  /* read all the files in the dir ~ */
+  vector<string> res;
+  while ((filename = readdir(dir)) != nullptr) {
+    string d_name = string(filename->d_name);
+    // get rid of "." and ".."
+    if (d_name == "." || d_name == ".." || filename->d_type != DT_REG)
+      continue;
+    res.emplace_back(string(dir_name) + "/" + filename->d_name);
+  }
+
+  sort(res.begin(), res.end());
+  return res;
+}
+
+
+int WriteResult(const string& imageFile, const vector<MSTensor> &outputs) {
+    string homePath = "./result_Files/";
+    const int INVALID_POINTER = -1;
+    const int ERROR = -2;
+    for (size_t i = 0; i < outputs.size(); ++i) {
+        size_t outputSize;
+        shared_ptr<const void> netOutput = outputs[i].Data();
+        outputSize = outputs[i].DataSize();
+        int pos = imageFile.rfind('/');
+        string fileName(imageFile, pos + 1);
+        string outFileName = homePath + "/" + fileName;
+        FILE *outputFile = fopen(outFileName.c_str(), "wb");
+        if (outputFile == nullptr) {
+            cout << "open result file " << outFileName << " failed" << endl;
+            return INVALID_POINTER;
+        }
+        size_t size = fwrite(netOutput.get(), sizeof(char), outputSize, outputFile);
+        if (size != outputSize) {
+            fclose(outputFile);
+            outputFile = nullptr;
+            cout << "write result file " << outFileName << " failed, write size[" << size <<
+                "] is smaller than output size[" << outputSize << "], maybe the disk is full." << endl;
+            return ERROR;
+        }
+        fclose(outputFile);
+        outputFile = nullptr;
+    }
+    return 0;
+}
+
+
+
+
+
+DIR *OpenDir(string_view dir_name) {
+  // check the parameter !
+  if (dir_name.empty()) {
+    cout << " dir_name is null ! " << endl;
+    return nullptr;
+  }
+
+  string real_path = RealPath(dir_name);
+
+  // check if dir_name is a valid dir
+  struct stat s;
+  lstat(real_path.c_str(), &s);
+  if (!S_ISDIR(s.st_mode)) {
+    cout << "dir_name is not a valid directory !" << endl;
+    return nullptr;
+  }
+
+  DIR *dir;
+  dir = opendir(real_path.c_str());
+  if (dir == nullptr) {
+    cout << "Can not open dir " << dir_name << endl;
+    return nullptr;
+  }
+  return dir;
+}
+
+
+
+string RealPath(string_view path) {
+  char real_path_mem[PATH_MAX] = {0};
+  char *real_path_ret = realpath(path.data(), real_path_mem);
+
+  if (real_path_ret == nullptr) {
+    cout << "File: " << path << " is not exist.";
+    return "";
+  }
+
+  return string(real_path_mem);
+}
+
+
+
+ms::MSTensor ReadFile(const string &file) {
+  if (file.empty()) {
+    cout << "Pointer file is nullptr" << endl;
+    return ms::MSTensor();
+  }
+
+  ifstream ifs(file);
+  if (!ifs.good()) {
+    cout << "File: " << file << " is not exist" << endl;
+    return ms::MSTensor();
+  }
+
+  if (!ifs.is_open()) {
+    cout << "File: " << file << "open failed" << endl;
+    return ms::MSTensor();
+  }
+
+  ifs.seekg(0, ios::end);
+  size_t size = ifs.tellg();
+  ms::MSTensor buffer(file, ms::DataType::kNumberTypeUInt8, {static_cast<int64_t>(size)}, nullptr, size);
+
+  ifs.seekg(0, ios::beg);
+  ifs.read(reinterpret_cast<char *>(buffer.MutableData()), size);
+  ifs.close();
+
+  return buffer;
+}
+size_t GetMax(ms::MSTensor data) {
+  float max_value = -1;
+  size_t max_idx = 0;
+  const float *p = reinterpret_cast<const float *>(data.MutableData());
+  for (size_t i = 0; i < data.DataSize() / sizeof(float); ++i) {
+    if (p[i] > max_value) {
+      max_value = p[i];
+      max_idx = i;
+    }
+  }
+  return max_idx;
+}
diff --git a/research/cv/ProtoNet/export.py b/research/cv/ProtoNet/export.py
index 69e0ab363b753203639bd7c411d38d2fa46eadd9..8ec5f0b78aec755bf40858e2ea4519ef4ba9c04b 100644
--- a/research/cv/ProtoNet/export.py
+++ b/research/cv/ProtoNet/export.py
@@ -24,10 +24,10 @@ from src.protonet import ProtoNet as protonet
 
 parser = argparse.ArgumentParser(description='MindSpore MNIST Example')
 parser.add_argument("--device_id", type=int, default=0, help="Device id")
-parser.add_argument("--batch_size", type=int, default=1, help="batch size")
+parser.add_argument("--batch_size", type=int, default=100, help="batch size")
 parser.add_argument("--ckpt_file", type=str, required=True, help="Checkpoint file path.")
 parser.add_argument("--file_name", type=str, default="protonet", help="output file name.")
-parser.add_argument("--file_format", type=str, choices=["AIR", "ONNX", "MINDIR"], default="AIR", help="file format")
+parser.add_argument("--file_format", type=str, choices=["AIR", "ONNX", "MINDIR"], default="MINDIR", help="file format")
 parser.add_argument("--device_target", type=str, choices=["Ascend", "GPU", "CPU"], default="Ascend",
                     help="device target")
 args = parser.parse_args()
@@ -40,6 +40,7 @@ if __name__ == "__main__":
 
     # define fusion network
     network = protonet()
+
     # load network checkpoint
     param_dict = load_checkpoint(args.ckpt_file)
     load_param_into_net(network, param_dict)
diff --git a/research/cv/ProtoNet/loss_for_infer.py b/research/cv/ProtoNet/loss_for_infer.py
new file mode 100644
index 0000000000000000000000000000000000000000..721b5c056da0f565ceb1eee4642431b04b790313
--- /dev/null
+++ b/research/cv/ProtoNet/loss_for_infer.py
@@ -0,0 +1,102 @@
+# Copyright 2021 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.
+# ============================================================================
+"""
+loss function script.
+"""
+import heapq
+import numpy as np
+
+
+def calculate_loss(inp, target, classes, n_support, n_query, n_class, is_train=True):
+    """
+    loss construct
+    """
+    n_classes = len(classes)
+    support_idxs = ()
+    query_idxs = ()
+
+    for ind, _ in enumerate(classes):
+        class_c = classes[ind]
+        matrix = np.equal(target, class_c).astype(np.float32)
+        K = n_support + n_query
+        a = heapq.nlargest(K, range(len(matrix)), matrix.take)
+        support_idx = np.squeeze(a[:n_support])
+        support_idxs += (support_idx,)
+        query_idx = a[n_support:]
+        query_idxs += (query_idx,)
+
+    prototypes = ()
+    for idx_list in support_idxs:
+        prototypes += (np.mean(inp[idx_list], axis=0),)
+    prototypes = np.stack(prototypes)
+    query_idxs = np.stack(query_idxs).reshape(-1)
+    query_samples = inp[query_idxs]
+
+
+    dists = euclidean_dist(query_samples, prototypes)
+    log_p_y = np.log(np.exp(-dists) / np.sum(np.exp(-dists)))
+    log_p_y = log_p_y.reshape((n_classes, n_query, -1))
+
+    target_inds = np.arange(0, n_class, dtype=np.int32).reshape((n_classes, 1, 1))
+    target_inds = np.broadcast_to(target_inds, (n_classes, n_query, 1))
+    loss_val = -np.mean(np.squeeze(gather(log_p_y, 2, target_inds).reshape(-1)))
+
+    y_hat = np.argmax(log_p_y, axis=2)
+    acc_val = np.mean(np.equal(y_hat, np.squeeze(target_inds)).astype(np.float32))
+    if is_train:
+        return loss_val
+    return acc_val, loss_val
+
+def supp_idxs(target, c):
+    return np.squeeze(nonZero(np.equal(target, c))[:n_support])
+
+def nonZero(inpbool):
+    out = []
+    for _, inp in enumerate(inpbool):
+        if inp:
+            out.append(inp)
+    return np.array(out, dtype=np.int32)
+
+def acc():
+    return acc_val
+
+def gather(self, dim, index):
+    '''
+    gather
+    '''
+    idx_xsection_shape = index.shape[:dim] + \
+        index.shape[dim + 1:]
+    self_xsection_shape = self.shape[:dim] + self.shape[dim + 1:]
+    if idx_xsection_shape != self_xsection_shape:
+        raise ValueError("Except for dimension " + str(dim) +
+                         ", all dimensions of index and self should be the same size")
+    data_swaped = np.swapaxes(self, 0, dim)
+    index_swaped = np.swapaxes(index, 0, dim)
+    gathered = np.choose(index_swaped, data_swaped)
+    return np.swapaxes(gathered, 0, dim)
+
+def euclidean_dist(x, y):
+    '''
+    Compute euclidean distance between two tensors
+    '''
+    # x: N x D
+    # y: M x D
+    n = x.shape[0]
+    m = y.shape[0]
+    d = x.shape[1]
+    x = np.broadcast_to(np.expand_dims(x, axis=1), (n, m, d))
+    y = np.broadcast_to(np.expand_dims(y, axis=0), (n, m, d))
+
+    return np.sum(np.power(x - y, 2), 2)
diff --git a/research/cv/ProtoNet/postprocess.py b/research/cv/ProtoNet/postprocess.py
new file mode 100644
index 0000000000000000000000000000000000000000..e2b6fa6e6ab0cc581d4240b26910ec564479c582
--- /dev/null
+++ b/research/cv/ProtoNet/postprocess.py
@@ -0,0 +1,51 @@
+'''
+calculate the accuracy using the infer result which are binary files
+'''
+import os
+import argparse
+import numpy as np
+from loss_for_infer import calculate_loss
+
+parser = argparse.ArgumentParser()
+parser.add_argument('--result_path', default=None, help='Location of result.')
+parser.add_argument('--label_classses_path', default=None, help='Location of label and classes.')
+parser.add_argument('--classes_per_it_val', type=int,
+                    help='number of random classes per episode for validation, default=5', default=5)
+parser.add_argument('--num_support_val', type=int,
+                    help='number of samples per class to use as support for validation, default=5', default=5)
+parser.add_argument('--num_query_val', type=int,
+                    help='number of samples per class to use as query for validation, default=15', default=15)
+
+def get_result(options):
+    '''
+    calculate the acc
+    '''
+    files = os.listdir(options.result_path)
+    acc = list()
+    loss = list()
+    for file in files:
+        result_file_name = file
+        num = result_file_name.split('_')[1]
+
+        result_file_path = options.result_path + os.sep + result_file_name
+        label_file_path = options.label_classses_path + os.sep + 'label_' + str(num)
+        classes_file_path = options.label_classses_path + os.sep + 'classes_' + str(num)
+
+        output = np.fromfile(result_file_path, dtype=np.float32)
+        label = np.fromfile(label_file_path, dtype=np.int32)
+        classes = np.fromfile(classes_file_path, dtype=np.int32)
+        batch_size = (options.num_support_val + options.num_query_val) * options.classes_per_it_val
+        # 64 is the fixed output dimension of the model
+        output = np.reshape(output, (batch_size, 64))
+
+        acc_val, loss_val = calculate_loss(output, label, classes, options.num_support_val,
+                                           options.num_query_val, options.classes_per_it_val, is_train=False)
+        acc.append(acc_val)
+        loss.append(loss_val)
+    mean_acc = sum(acc) / len(acc)
+    mean_loss = sum(loss) / len(loss)
+    print("accuracy: {}  loss:{}".format(mean_acc, mean_loss))
+
+if __name__ == '__main__':
+    options_ = parser.parse_args()
+    get_result(options_)
diff --git a/research/cv/ProtoNet/preprocess.py b/research/cv/ProtoNet/preprocess.py
new file mode 100644
index 0000000000000000000000000000000000000000..813c33e5cb5d959eae9b17bc9398d1660291942d
--- /dev/null
+++ b/research/cv/ProtoNet/preprocess.py
@@ -0,0 +1,50 @@
+'''
+preprocess the source data and generate the result data with binary file
+'''
+import os
+import argparse
+from model_init import init_dataloader
+from mindspore import dataset as ds
+
+
+parser = argparse.ArgumentParser()
+parser.add_argument('--dataset_path', default=None, help='Location of data.')
+parser.add_argument('--data_output_path', default=None, help='Location of converted data.')
+parser.add_argument('--label_classses_output_path', default=None,
+                    help='Location of converted label and classes.')
+parser.add_argument('-its', '--iterations', type=int, help='number of episodes per epoch, default=100',
+                    default=100)
+parser.add_argument('-cTr', '--classes_per_it_tr', type=int,
+                    help='number of random classes per episode for training, default=60', default=20)
+parser.add_argument('-nsTr', '--num_support_tr', type=int,
+                    help='number of samples per class to use as support for training, default=5', default=5)
+parser.add_argument('-nqTr', '--num_query_tr', type=int,
+                    help='number of samples per class to use as query for training, default=5', default=5)
+parser.add_argument('-cVa', '--classes_per_it_val', type=int,
+                    help='number of random classes per episode for validation, default=5', default=5)
+parser.add_argument('-nsVa', '--num_support_val', type=int,
+                    help='number of samples per class to use as support for validation, default=5', default=5)
+parser.add_argument('-nqVa', '--num_query_val', type=int,
+                    help='number of samples per class to use as query for validation, default=15', default=15)
+
+def convert_img_to_bin(options_, root, output_path, label_classses_path):
+    '''
+    convert the image to binary file
+    '''
+    val_dataloader = init_dataloader(options_, 'val', root)
+    inp = ds.GeneratorDataset(val_dataloader, column_names=['data', 'label', 'classes'])
+    i = 1
+    for batch in inp.create_dict_iterator():
+        x = batch['data']
+        y = batch['label']
+        classes = batch['classes']
+        x_array = x.asnumpy()
+        y_array = y.asnumpy()
+        classes_array = classes.asnumpy()
+        x_array.tofile(output_path + os.sep +"data_" + str(i) + ".bin")
+        y_array.tofile(label_classses_path + os.sep +"label_" + str(i) + ".bin")
+        classes_array.tofile(label_classses_path + os.sep +"classes_" + str(i) + ".bin")
+        i = i + 1
+if __name__ == '__main__':
+    options = parser.parse_args()
+    convert_img_to_bin(options, options.dataset_path, options.data_output_path, options.label_classses_output_path)
diff --git a/research/cv/ProtoNet/scripts/run_infer_310.sh b/research/cv/ProtoNet/scripts/run_infer_310.sh
new file mode 100644
index 0000000000000000000000000000000000000000..ed6c5cf78405076de14b3ddffe62ccb88cbfffb6
--- /dev/null
+++ b/research/cv/ProtoNet/scripts/run_infer_310.sh
@@ -0,0 +1,113 @@
+#!/bin/bash
+# Copyright 2021 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 2 || $# -gt 3 ]]; then
+    echo "Usage: bash run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [DEVICE_ID]
+    DEVICE_ID is optional, it can be set by environment variable device_id, otherwise the value is zero"
+exit 1
+fi
+
+get_real_path(){
+    if [ "${1:0:1}" == "/" ]; then
+        echo "$1"
+    else
+        echo "$(realpath -m $PWD/$1)"
+    fi
+}
+
+model=$(get_real_path $1)
+data_path=$(get_real_path $2)
+device_id=0
+if [ $# == 3 ]; then
+    device_id=$3
+fi
+
+echo "mindir name: "$model
+echo "dataset path: "$data_path
+echo "device id: "$device_id
+
+export ASCEND_HOME=/usr/local/Ascend/
+if [ -d ${ASCEND_HOME}/ascend-toolkit ]; then
+    export PATH=$ASCEND_HOME/fwkacllib/bin:$ASCEND_HOME/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/ascend-toolkit/latest/atc/bin:$PATH
+    export LD_LIBRARY_PATH=$ASCEND_HOME/fwkacllib/lib64:/usr/local/lib:$ASCEND_HOME/ascend-toolkit/latest/atc/lib64:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH
+    export TBE_IMPL_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp/op_impl/built-in/ai_core/tbe
+    export PYTHONPATH=$ASCEND_HOME/fwkacllib/python/site-packages:${TBE_IMPL_PATH}:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/python/site-packages:$PYTHONPATH
+    export ASCEND_OPP_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp
+else
+    export PATH=$ASCEND_HOME/fwkacllib/bin:$ASCEND_HOME/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/atc/ccec_compiler/bin:$ASCEND_HOME/atc/bin:$PATH
+    export LD_LIBRARY_PATH=$ASCEND_HOME/fwkacllib/lib64:/usr/local/lib:$ASCEND_HOME/atc/lib64:$ASCEND_HOME/acllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH
+    export PYTHONPATH=$ASCEND_HOME/fwkacllib/python/site-packages:$ASCEND_HOME/atc/python/site-packages:$PYTHONPATH
+    export ASCEND_OPP_PATH=$ASCEND_HOME/opp
+fi
+
+function preprocess_data()
+{
+    if [ -d data_preprocess_Result ]; then
+        rm -rf ./data_preprocess_Result
+    fi
+    if [ -d label_classes_preprocess_Result ]; then
+        rm -rf ./label_classes_preprocess_Result
+    fi
+    mkdir data_preprocess_Result
+    mkdir label_classes_preprocess_Result
+    python ../preprocess.py --dataset_path=$data_path --data_output_path=./data_preprocess_Result --label_classses_output_path=./label_classes_preprocess_Result &> preprocess.log
+    data_path=./data_preprocess_Result
+    label_classes_path=./label_classes_preprocess_Result/
+}
+
+function compile_app()
+{
+    cd ../ascend310_infer || exit
+    bash build.sh &> build.log
+}
+
+function infer()
+{
+    cd - || exit
+    if [ -d result_Files ]; then
+        rm -rf ./result_Files
+    fi
+    mkdir result_Files
+    ../ascend310_infer/out/main --mindir_path=$model --dataset_path=$data_path --device_id=$device_id &> infer.log
+}
+
+function cal_acc()
+{
+    python3.7 ../postprocess.py --result_path=./result_Files/ --label_classses_path=$label_classes_path &> acc.log &
+}
+
+
+
+preprocess_data
+if [ $? -ne 0 ]; then
+    echo "preprocess data failed"
+    exit 1
+fi
+compile_app
+if [ $? -ne 0 ]; then
+    echo "compile app code failed"
+    exit 1
+fi
+infer
+if [ $? -ne 0 ]; then
+    echo " execute inference failed"
+    exit 1
+fi
+cal_acc
+if [ $? -ne 0 ]; then
+    echo "calculate accuracy failed"
+    exit 1
+fi
\ No newline at end of file