diff --git a/mindspore/ccsrc/minddata/dataset/api/vision.cc b/mindspore/ccsrc/minddata/dataset/api/vision.cc
index e63dc87343..afffb47fa4 100644
--- a/mindspore/ccsrc/minddata/dataset/api/vision.cc
+++ b/mindspore/ccsrc/minddata/dataset/api/vision.cc
@@ -691,14 +691,14 @@ Status DvppDecodeResizeCropOperation::ValidateParams() {
     RETURN_STATUS_SYNTAX_ERROR(err_msg);
   }
   if (crop_.size() < resize_.size()) {
-    if (crop_[0] >= MIN(resize_[0], resize_[1])) {
+    if (crop_[0] > MIN(resize_[0], resize_[1])) {
       std::string err_msg = "crop size must be smaller than resize size";
       MS_LOG(ERROR) << err_msg;
       RETURN_STATUS_SYNTAX_ERROR(err_msg);
     }
   }
   if (crop_.size() > resize_.size()) {
-    if (MAX(crop_[0], crop_[1]) >= resize_[0]) {
+    if (MAX(crop_[0], crop_[1]) > resize_[0]) {
       std::string err_msg = "crop size must be smaller than resize size";
       MS_LOG(ERROR) << err_msg;
       RETURN_STATUS_SYNTAX_ERROR(err_msg);
@@ -706,7 +706,7 @@ Status DvppDecodeResizeCropOperation::ValidateParams() {
   }
   if (crop_.size() == resize_.size()) {
     for (int32_t i = 0; i < crop_.size(); ++i) {
-      if (crop_[i] >= resize_[i]) {
+      if (crop_[i] > resize_[i]) {
         std::string err_msg = "crop size must be smaller than resize size";
         MS_LOG(ERROR) << err_msg;
         RETURN_STATUS_SYNTAX_ERROR(err_msg);
diff --git a/tests/cxx_st/dataset/test_de.cc b/tests/cxx_st/dataset/test_de.cc
index 2d0aaf923f..510324867e 100644
--- a/tests/cxx_st/dataset/test_de.cc
+++ b/tests/cxx_st/dataset/test_de.cc
@@ -20,10 +20,31 @@
 #include "minddata/dataset/include/minddata_eager.h"
 #include "minddata/dataset/include/vision.h"
 #include "minddata/dataset/kernels/tensor_op.h"
+#include "include/api/model.h"
+#include "include/api/serializations.h"
+#include "include/api/context.h"
 
 using namespace mindspore::api;
 using namespace mindspore::dataset::vision;
 
+static void SaveFile(int idx, Buffer buffer, int seq) {
+  std::string path = "mnt/disk1/yolo_dvpp_result/result_Files/output" + std::to_string(idx) +
+                     "_in_YoloV3-DarkNet_coco_bs_dvpp_" + std::to_string(seq) + ".bin";
+  FILE *output_file = fopen(path.c_str(), "wb");
+  if (output_file == nullptr) {
+    std::cout << "Write file" << path << "failed when fopen" << std::endl;
+    return;
+  }
+
+  size_t wsize = fwrite(buffer.Data(), buffer.DataSize(), sizeof(int8_t), output_file);
+  if (wsize == 0) {
+    std::cout << "Write file" << path << " failed when fwrite." << std::endl;
+    return;
+  }
+  fclose(output_file);
+  std::cout << "Save file " << path << "length" << buffer.DataSize() << " success." << std::endl;
+}
+
 class TestDE : public ST::Common {
  public:
   TestDE() {}
@@ -56,10 +77,44 @@ TEST_F(TestDE, TestDvpp) {
 
   for (auto &img : images) {
     img = Solo(img);
+    ASSERT_EQ(images[0]->Shape().size(), 3);
+    ASSERT_EQ(images[0]->Shape()[0], 224 * 224 * 1.5);
+    ASSERT_EQ(images[0]->Shape()[1], 1);
+    ASSERT_EQ(images[0]->Shape()[2], 1);
   }
+}
 
-  ASSERT_EQ(images[0]->Shape().size(), 3);
-  ASSERT_EQ(images[0]->Shape()[0], 224 * 224 * 1.5);
-  ASSERT_EQ(images[0]->Shape()[1], 1);
-  ASSERT_EQ(images[0]->Shape()[2], 1);
+TEST_F(TestDE, TestYoloV3_with_Dvpp) {
+  std::vector<std::shared_ptr<Tensor>> images;
+  MIndDataEager::LoadImageFromDir("/home/lizhenglong/val2014", &images);
+  MindDataEager SingleOp({DvppDecodeResizeCropJpeg({416, 416}, {416, 416})});
+  constexpr auto yolo_mindir_file = "/home/zhoufeng/yolov3/yolov3_darknet53.mindir";
+  Context::Instance().SetDeviceTarget(kDeviceTypeAscend310).SetDeviceID(1);
+  auto graph = Serialization::LoadModel(yolo_mindir_file, ModelType::kMindIR);
+  Model yolov3((GraphCell(graph)));
+  Status ret = yolov3.Build({{kMOdelOptionInsertOpCfgPath, "/mnt/disk1/yolo_dvpp_result/aipp_resnet50.cfg"}});
+  ASSERT_TRUE(ret == SUCCESS);
+
+  std::vector<std::string> names;
+  std::vector<std::vector<int64_t>> shapes;
+  std::vector<DataType> data_types;
+  std::vector<size_t> mem_sizes;
+  yolov3.GetOutputsInfo(&names, &shapes, &data_types, &mem_sizes);
+  std::vector<Buffer> outputs;
+  std::vector<Buffer> inputs;
+
+  int64_t seq = 0;
+  for (auto &img : images) {
+    img = SingleOp(img);
+    std::vector<float> input_shape = {416, 416};
+    input.clear();
+    inputs.emplace_back(img->data(), img->DataSize());
+    inputs.emplace_back(input_shape.data(), input_shape.size() * sizeof(float));
+    ret = yolov3.Predict(inputs, &outputs);
+    for (size_t i = 0; i < outputs.size(); ++i) {
+      SaveFile(i, outputs[i], seq);
+    }
+    seq++;
+    ASSERT_TRUE(ret == SUCCESS);
+  }
 }