diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/CMakeLists.txt b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/CMakeLists.txt
index 617c5dd39858f6df17f1d97e96ec09af3012ae42..4685de84747af36e235c37e9632adbdeb7b62613 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/CMakeLists.txt
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/CMakeLists.txt
@@ -47,5 +47,6 @@ link_directories(${MXBASE_POST_LIB_DIR})
 
 add_executable(${TARGET} src/main.cpp src/SpansNet100.cpp)
 target_link_libraries(${TARGET} glog cpprest mxbase opencv_world stdc++fs)
+target_link_libraries(${TARGET} glog cpprest mxbase resnet50postprocess opencv_world stdc++fs)
 
 install(TARGETS ${TARGET} RUNTIME DESTINATION ${PROJECT_SOURCE_DIR}/)
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/build.sh b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/build.sh
index a4ad292b21cb703b2d7d4e1d9fcf4b2f898078d1..7514c9ffc2a15243e0322f674163f78dc6ead04b 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/build.sh
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/build.sh
@@ -36,7 +36,11 @@ function check_env() {
 function build_spansnet100() {
     cd $path_cur
     rm -rf build
+    if [ -d "output" ]; then
+      rm -rf output
+    fi
     mkdir -p build
+    mkdir -p output
     cd build
     cmake ..
     make
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.cpp b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.cpp
index b6de70f0bc42a92578318cbecb1fe88fdbcad6c7..d50e27ed5ecf954dbe90401f42cd5119ac92c8af 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.cpp
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.cpp
@@ -23,6 +23,7 @@
 #include <vector>
 #include "MxBase/DeviceManager/DeviceManager.h"
 #include "MxBase/Log/Log.h"
+#include <opencv2/dnn.hpp>
 
 APP_ERROR SpansNet100Infer::Init(const InitParam &initParam) {
     deviceId_ = initParam.deviceId;
@@ -42,6 +43,26 @@ APP_ERROR SpansNet100Infer::Init(const InitParam &initParam) {
         LogError << "ModelInferenceProcessor init failed, ret=" << ret << ".";
         return ret;
     }
+    MxBase::ConfigData configData;
+    const std::string softmax = initParam.softmax ? "true" : "false";
+    const std::string checkTensor = initParam.checkTensor ? "true" : "false";
+
+    configData.SetJsonValue("CLASS_NUM", std::to_string(initParam.classNum));
+    configData.SetJsonValue("TOP_K", std::to_string(initParam.topk));
+    configData.SetJsonValue("SOFTMAX", softmax);
+    configData.SetJsonValue("CHECK_MODEL", checkTensor);
+
+    auto jsonStr = configData.GetCfgJson().serialize();
+    std::map<std::string, std::shared_ptr<void>> config;
+    config["postProcessConfigContent"] = std::make_shared<std::string>(jsonStr);
+    config["labelPath"] = std::make_shared<std::string>(initParam.labelPath);
+
+    post_ = std::make_shared<MxBase::Resnet50PostProcess>();
+    ret = post_->Init(config);
+    if (ret != APP_ERR_OK) {
+        LogError << "Resnet50PostProcess init failed, ret=" << ret << ".";
+        return ret;
+    }
     return APP_ERR_OK;
 }
 
@@ -77,6 +98,7 @@ APP_ERROR SpansNet100Infer::BinToTensorBase(const std::string file_path, MxBase:
 
 APP_ERROR SpansNet100Infer::DeInit() {
     model_->DeInit();
+    post_->DeInit();
     MxBase::DeviceManager::GetInstance()->DestroyDevices();
     return APP_ERR_OK;
 }
@@ -112,49 +134,45 @@ APP_ERROR SpansNet100Infer::Inference(const std::vector<MxBase::TensorBase> &inp
     return APP_ERR_OK;
 }
 
-APP_ERROR SpansNet100Infer::PostProcess(std::vector<MxBase::TensorBase> *outputs,
-                                        std::vector<float> *predict) {
-    MxBase::TensorBase &tensor = outputs->at(0);
-    APP_ERROR ret = tensor.ToHost();
+APP_ERROR SpansNet100Infer::PostProcess(const std::vector<MxBase::TensorBase> &inputs,
+                                        std::vector<std::vector<MxBase::ClassInfo>> &clsInfos)
+{
+    APP_ERROR ret = post_->Process(inputs, clsInfos);
     if (ret != APP_ERR_OK) {
-        LogError << GetError(ret) << "Tensor deploy to host failed.";
+        LogError << "Process failed, ret=" << ret << ".";
         return ret;
     }
-    // check tensor is available
-    auto outputShape = tensor.GetShape();
-    uint32_t length = outputShape[0] * outputShape[1];
-    void *data = tensor.GetBuffer();
-    for (uint32_t i = 0; i < length; i++) {
-        float value = *(reinterpret_cast<float *>(data) + i);
-        predict->push_back(value);
-    }
     return APP_ERR_OK;
 }
 
-APP_ERROR SpansNet100Infer::WriteResult(const std::string &fileName, const std::vector<float> &predict,
-                                        const std::string name) {
+APP_ERROR SpansNet100Infer::SaveResult(const std::vector<std::vector<MxBase::ClassInfo>> &batchClsInfos, const std::string name)
+{
     // create result file
-    std::ofstream tfile(fileName, std::ofstream::app);
+    std::string file_name = "./output/" + name + ".txt";
+    std::ofstream tfile(file_name, std::ofstream::app);
     if (tfile.fail()) {
-        LogError << "Failed to open result file: " << fileName;
+        LogError << "Failed to open result file: " << file_name;
         return APP_ERR_COMM_OPEN_FAIL;
     }
     // write inference result into file
     LogInfo << "==============================================================";
     LogInfo << "Infer finished!";
-
-    tfile<<name;
-    for (uint32_t i = 0; i < predict.size(); i++){
-        tfile << " " << predict[i];
+    tfile << name << " ";
+    for (auto clsInfos : batchClsInfos) {
+        std::string resultStr;
+        for (auto clsInfo : clsInfos) {
+            LogDebug << " className:" << clsInfo.className << " confidence:" << clsInfo.confidence <<
+                     " classIndex:" <<  clsInfo.classId;
+            resultStr += std::to_string(clsInfo.classId) + " ";
+        }
+        tfile << resultStr << std::endl;
     }
-    tfile << std::endl;
-
     LogInfo << "==============================================================";
     tfile.close();
     return APP_ERR_OK;
 }
 
-APP_ERROR SpansNet100Infer::Process(const std::string &imgPath, const std::string &outputPath) {
+APP_ERROR SpansNet100Infer::Process(const std::string &imgPath) {
     MxBase::TensorBase tensorBase;
     APP_ERROR ret = BinToTensorBase(imgPath, tensorBase);
     if (ret != APP_ERR_OK) {
@@ -166,9 +184,9 @@ APP_ERROR SpansNet100Infer::Process(const std::string &imgPath, const std::strin
     std::vector<MxBase::TensorBase> outputs = {};
     inputs.push_back(tensorBase);
     ret = Inference(inputs, outputs);
-    // output to vector
-    std::vector<float> predict;
-    ret = PostProcess(&outputs, &predict);
+    // postprocess
+    std::vector<std::vector<MxBase::ClassInfo>> batchClsInfos = {};
+    ret = PostProcess(outputs, batchClsInfos);
     if (ret != APP_ERR_OK) {
         LogError << "PostProcess failed, ret=" << ret << ".";
         return ret;
@@ -177,9 +195,9 @@ APP_ERROR SpansNet100Infer::Process(const std::string &imgPath, const std::strin
     int index = imgPath.find_last_of("/");
     std::string name = imgPath.substr(index + 1, 23);
     // save result
-    ret = WriteResult(outputPath, predict, name);
+    ret = SaveResult(batchClsInfos, name);
     if (ret != APP_ERR_OK) {
-        LogError << "save result failed, ret=" << ret << ".";
+        LogError << "Save infer results into file failed. ret = " << ret << ".";
         return ret;
     }
     return APP_ERR_OK;
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.h b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.h
index 03b1b45b1176297d65c20dc748edb02aa9500f27..cf28ea867936693069270f08e6233c864dab6752 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.h
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/SpansNet100.h
@@ -24,6 +24,7 @@
 #include <opencv2/opencv.hpp>
 #include "MxBase/DvppWrapper/DvppWrapper.h"
 #include "MxBase/ModelInfer/ModelInferenceProcessor.h"
+#include "ClassPostProcessors/Resnet50PostProcess.h"
 #include "MxBase/Tensor/TensorContext/TensorContext.h"
 
 extern std::vector<double> g_inferCost;
@@ -31,7 +32,10 @@ extern std::vector<double> g_inferCost;
 struct InitParam {
     uint32_t deviceId;
     std::string labelPath;
-    uint32_t topk=5;
+    uint32_t classNum;
+    uint32_t topk;
+    bool softmax;
+    bool checkTensor;
     std::string modelPath;
 };
 
@@ -42,17 +46,15 @@ public:
     APP_ERROR BinToTensorBase(const std::string file_path, MxBase::TensorBase &tensorBase);
     APP_ERROR Inference(const std::vector<MxBase::TensorBase> &inputs,
                         std::vector<MxBase::TensorBase> &outputs);
-    APP_ERROR PostProcess(std::vector<MxBase::TensorBase> *outputs,
-                          std::vector<float> *predict);
-    APP_ERROR Process(const std::string &imgPath,
-                      const std::string &outputPath);
-    APP_ERROR WriteResult(const std::string &fileName,
-                          const std::vector<float> &predict,
-                          const std::string name);
+    APP_ERROR PostProcess(const std::vector<MxBase::TensorBase> &inputs,
+                          std::vector<std::vector<MxBase::ClassInfo>> &clsInfos);
+    APP_ERROR Process(const std::string &imgPath);
+    APP_ERROR SaveResult(const std::vector<std::vector<MxBase::ClassInfo>> &batchClsInfos, const std::string name);
 
 private:
     std::shared_ptr<MxBase::DvppWrapper> dvppWrapper_;
     std::shared_ptr<MxBase::ModelInferenceProcessor> model_;
+    std::shared_ptr<MxBase::Resnet50PostProcess> post_;
     MxBase::ModelDesc modelDesc_;
     uint32_t deviceId_ = 0;
 };
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/main.cpp b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/main.cpp
index 8229156abe964763a9c119fa5989d49f8a692776..435c97b3c8e53f07def479854deb84f3ed543557 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/main.cpp
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/src/main.cpp
@@ -22,15 +22,24 @@
 
 namespace fs = std::experimental::filesystem;
 std::vector<double> g_inferCost;
+namespace {
+    const uint32_t CLASS_NUM = 1000;
+} // namespace
 
 int main(int argc, char *argv[]) {
     if (argc <= 1) {
-        LogWarn << "Please input image path and ouput path, such as '../data/input/preprocess ../data/input/preprocess_result.txt";
+        LogWarn << "Please input image path and ouput path, such as '../data/input/preprocess";
         return APP_ERR_OK;
     }
     InitParam initParam = {};
     initParam.deviceId = 0;
+    initParam.classNum = CLASS_NUM;
+    initParam.topk=5;
+    initParam.softmax = false;
+    initParam.checkTensor = true;
+
     initParam.modelPath = "../data/model/spnasnet_100_bs1.om";
+    initParam.labelPath = "../data/input/imagenet1000_clsidx_to_labels.txt";
     auto spansnet = std::make_shared<SpansNet100Infer>();
     APP_ERROR ret = spansnet->Init(initParam);
     if (ret != APP_ERR_OK) {
@@ -38,12 +47,11 @@ int main(int argc, char *argv[]) {
         return ret;
     }
     std::string imgDir = argv[1];
-    std::string outputDir = argv[2];
     int index = 0;
     for (auto &entry : fs::directory_iterator(imgDir)) {
         index++;
         LogInfo << "read image path " << entry.path();
-        ret = spansnet->Process(entry.path(), outputDir);
+        ret = spansnet->Process(entry.path());
         if (ret != APP_ERR_OK) {
             LogError << "SpansNet100Infer process failed, ret=" << ret << ".";
             spansnet->DeInit();
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/utils/eval.py b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/utils/eval.py
index 66e0007872229565f27cceb7756c236ea437bb30..c9a169d32c3947b1f2f609c23c4f9c843fc374b1 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/utils/eval.py
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/mxbase/utils/eval.py
@@ -17,13 +17,6 @@ import sys
 import os
 import json
 
-def w2txt(file_path, filename_list, data):
-    with open(file_path, "w") as file:
-        for i in range(data.shape[0]):
-            s = filename_list[i].split('.')[0] + ' '
-            s += ' '.join(str(int(num)) for num in data[i])
-            file.write(s+"\n")
-
 def cre_groundtruth_dict_fromtxt(gtfile_path):
     """
     :param filename: file contains the imagename and label number
@@ -53,54 +46,42 @@ def run():
     table_dict["value"] = []
     count = 0
     filename_list = []
-    sort_index_sents = []
-    with open(infer_reault, 'r')as f:
-        for line in f.readlines():
-            prediction = []
-            temp = line.strip().split(" ")
-            img_name = temp[0]
-            filename_list.append(img_name.split('.')[0])
-            print(img_name, "====", count)
-            count += 1
-            prediction = np.array([float(temp[i]) for i in range(1, len(temp))])
-            n_labels = len(prediction)
-            gt = img_gt_dict[img_name]
-            sort_index = np.argsort(-prediction)
-            if (n_labels == 1000):
-                realLabel = int(gt)
-            elif (n_labels == 1001):
-                realLabel = int(gt) + 1
-            else:
+    file_list = os.listdir(infer_reault)
+    for file in file_list:
+        with open(os.path.join(infer_reault, file), 'r')as f:
+            for line in f.readlines():
+                temp = line.strip().split(" ")
+                img_name = temp[0]
+                filename_list.append(img_name.split('.')[0])
+                print(img_name, "====", count)
+                count += 1
+                prediction = np.array([int(temp[i]) for i in range(1, len(temp))])
+                n_labels = len(prediction)
+                gt = img_gt_dict[img_name]
                 realLabel = int(gt)
-            resCnt = min(len(sort_index), topn)
-            for i in range(resCnt):
-                if (str(realLabel) == str(sort_index[i])):
-                    count_hit[i] += 1
-                    break
-            sort_index_sents.append(sort_index[:5])
-        file_path = "../mxbase_postprocess_result.txt"
-        sort_index_sents = np.array(sort_index_sents).astype(np.int32)
-        if(os.path.exists(file_path)):
-            os.remove(file_path)
-        w2txt(file_path, filename_list, sort_index_sents)
+                resCnt = topn
+                for i in range(resCnt):
+                    if (str(realLabel) == str(prediction[i])):
+                        count_hit[i] += 1
+                        break
 
-        print(count)
-        if 'value' not in table_dict.keys():
-            print("the item value does not exist!")
+    print(count)
+    if 'value' not in table_dict.keys():
+        print("the item value does not exist!")
+    else:
+        table_dict["value"].extend(
+            [{"key": "Number of images", "value": str(count)},
+             {"key": "Number of classes", "value": str(n_labels)}])
+        if count == 0:
+            accuracy = 0
         else:
-            table_dict["value"].extend(
-                [{"key": "Number of images", "value": str(count)},
-                 {"key": "Number of classes", "value": str(n_labels)}])
-            if count == 0:
-                accuracy = 0
-            else:
-                accuracy = np.cumsum(count_hit) / count
-            for i in range(resCnt):
-                table_dict["value"].append({"key": "Top" + str(i + 1) + " accuracy",
-                                            "value": str(
-                                                round(accuracy[i] * 100, 2)) + '%'})
-            json.dump(table_dict, writer)
-        writer.close()
+            accuracy = np.cumsum(count_hit) / count
+        for i in range(resCnt):
+            table_dict["value"].append({"key": "Top" + str(i + 1) + " accuracy",
+                                        "value": str(
+                                            round(accuracy[i] * 100, 2)) + '%'})
+        json.dump(table_dict, writer)
+    writer.close()
 
 
 if __name__ == '__main__':
diff --git a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/sdk/main.py b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/sdk/main.py
index 9eaa70eba03e57950f48bf08f68be76cecf16207..44e8eff439ceff6f5e9c0f40187beb086d190bcb 100644
--- a/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/sdk/main.py
+++ b/PyTorch/contrib/cv/classification/SPNASNet_100_for_PyTorch/infer/sdk/main.py
@@ -18,6 +18,7 @@
 import os
 import json
 import sys
+import shutil
 import time
 import MxpiDataType_pb2 as MxpiDataType
 import numpy as np
@@ -26,12 +27,11 @@ from StreamManagerApi import StreamManagerApi, InProtobufVector, \
     MxProtobufIn, StringVector
 
 
-def w2txt(file_path, filename_list, data):
-    with open(file_path, "w") as file:
-        for i in range(data.shape[0]):
-            s = filename_list[i].split('.')[0] + ' '
-            s += ' '.join(str(int(num)) for num in data[i])
-            file.write(s+"\n")
+def w2txt(file_path, filename_name, data):
+    with open(os.path.join(file_path, filename_name + ".txt"), "w") as file:
+        s = filename_name.split('.')[0] + ' '
+        s += ' '.join(str(int(num)) for num in data)
+        file.write(s+"\n")
 
 
 def send_source_data(appsrc_id, tensor, stream_name, stream_manager):
@@ -176,8 +176,11 @@ def run():
     count = 0
     resCnt = 0
     n_labels = 0
-    filename_list = []
-    sort_index_sents = []
+    out_path = "output"
+    if(os.path.exists(out_path)):
+        shutil.rmtree(out_path)
+    else:
+        os.mkdir(out_path)
     for file in file_list:
         print(file, "====", count)
         count += 1
@@ -214,7 +217,6 @@ def run():
         result = MxpiDataType.MxpiTensorPackageList()
         result.ParseFromString(infer_result[0].messageBuf)
         res = np.frombuffer(result.tensorPackageVec[0].tensorVec[0].dataStr, dtype='<f4')
-        filename_list.append(file.split('.')[0])
 
         # postprocess
         prediction = res
@@ -234,12 +236,9 @@ def run():
             if (str(realLabel) == str(sort_index[i])):
                 count_hit[i] += 1
                 break
-        sort_index_sents.append(sort_index[:5])
-    file_path = "sdk_postprocess_result.txt"
-    sort_index_sents = np.array(sort_index_sents).astype(np.int32)
-    if(os.path.exists(file_path)):
-        os.remove(file_path)
-    w2txt(file_path, filename_list, sort_index_sents)
+        data = np.array(sort_index[:5]).astype(np.int32)
+        file_name = file.split('.')[0]
+        w2txt(out_path, file_name, data)
 
     if 'value' not in table_dict.keys():
         print("the item value does not exist!")