!21685 add function to invert_op

Merge pull request !21685 from Killer/fake_killer

mindspore/ccsrc/minddata/dataset/kernels/image/invert_op.cc

@@ -32,22 +32,27 @@ Status InvertOp::Compute(const std::shared_ptr<Tensor> &input, std::shared_ptr<T
     if (!input_cv->mat().data) {
       RETURN_STATUS_UNEXPECTED("Invert: load image failed.");
     }
-
-    if (input_cv->Rank() != 3) {
-      RETURN_STATUS_UNEXPECTED("Invert: image shape is not <H,W,C>");
+    if (input_cv->Rank() != 3 && input_cv->Rank() != 2) {
+      RETURN_STATUS_UNEXPECTED("Invert: image shape is not <H,W,C> or <H,W>");
+    }
+    int num_channels = 1;
+    if (input_cv->Rank() == 3) {
+      num_channels = input_cv->shape()[2];
     }
-    int num_channels = input_cv->shape()[2];
-    if (num_channels != 3) {
-      RETURN_STATUS_UNEXPECTED(
-        "Invert: image shape is incorrect, expected num of channels is 3, "
-        "but got:" +
+    if (num_channels != 3 && num_channels != 1) {
+      RETURN_STATUS_UNEXPECTED("Invert: image shape is incorrect, expected num of channels is 1 or 3, but got:" + 
         std::to_string(num_channels));
     }
     std::shared_ptr<CVTensor> output_cv;
     RETURN_IF_NOT_OK(CVTensor::CreateEmpty(input_cv->shape(), input_cv->type(), &output_cv));
     RETURN_UNEXPECTED_IF_NULL(output_cv);
 
-    output_cv->mat() = cv::Scalar::all(255) - input_img;
+    if(num_channels == 3){
+      output_cv->mat() = cv::Scalar::all(255) - input_img;
+    }
+    else{
+      output_cv->mat() = cv::Scalar(255) - input_img;
+    }
     *output = std::static_pointer_cast<Tensor>(output_cv);
   }
 
@@ -58,3 +63,4 @@ Status InvertOp::Compute(const std::shared_ptr<Tensor> &input, std::shared_ptr<T
 }
 }  // namespace dataset
 }  // namespace mindspore
+

tests/ut/python/dataset/test_invert.py

@@ -25,6 +25,8 @@ from mindspore import log as logger
 from util import visualize_list, save_and_check_md5, diff_mse
 
 DATA_DIR = "../data/dataset/testImageNetData/train/"
+DATA_DIR0 = "../data/dataset/testImageNetData/train0/"
+DATA_DIR1 = "../data/dataset/testImageNetData/train1/"
 
 GENERATE_GOLDEN = False
 
@@ -201,8 +203,7 @@ def test_invert_py_c(plot=False):
     if plot:
         visualize_list(images_c_invert, images_p_invert, visualize_mode=2)
 
-
-def test_invert_one_channel():
+def test_invert_one_channel(plot=False):
     """
      Test Invert cpp op with one channel image
      """
@@ -211,17 +212,51 @@ def test_invert_one_channel():
     c_op = C.Invert()
 
     try:
-        data_set = ds.ImageFolderDataset(dataset_dir=DATA_DIR, shuffle=False)
-        data_set = data_set.map(operations=[C.Decode(), C.Resize((224, 224)),
+        data_set0 = ds.ImageFolderDataset(dataset_dir=DATA_DIR0, shuffle=False)
+        data_set0 = data_set0.map(operations=[C.Decode(), C.Resize((224, 224)),
+                                            lambda img: np.array(img[:, :, 0])], input_columns=["image"])
+
+        data_set1 = ds.ImageFolderDataset(dataset_dir=DATA_DIR1, shuffle=False)
+        data_set1 = data_set1.map(operations=[C.Decode(), C.Resize((224, 224)),
                                             lambda img: np.array(img[:, :, 0])], input_columns=["image"])
 
-        data_set.map(operations=c_op, input_columns="image")
+        data_uninvert = data_set0
+        data_uninvert = data_uninvert.batch(512)
+
+        for idx, (image, _) in enumerate(data_uninvert):
+            if idx == 0:
+                images_0_invert = image.asnumpy()
+            else:
+                images_0_invert = np.append(images_0_invert,
+                                            image.asnumpy(),
+                                            axis=0)
+
+        data_invert1 = data_set1.map(operations=c_op, input_columns="image")
+        data_invert1 = data_invert1.batch(512)
+
+        for idx, (image, _) in enumerate(data_invert1):
+            if idx == 0:
+                images_1_invert = image.asnumpy()
+            else:
+                images_1_invert = np.append(images_1_invert,
+                                            image.asnumpy(),
+                                            axis=0)
+	
+        num_samples = images_0_invert.shape[0]
+        mse = np.zeros(num_samples)
+        for i in range(num_samples):
+            mse[i] = diff_mse(images_0_invert[i], images_1_invert[i])
+        logger.info("MSE= {}".format(str(np.mean(mse))))
+
+        if plot:
+            visualize_list(images_0_invert, images_1_invert, visualize_mode=2)
 
     except RuntimeError as e:
         logger.info("Got an exception in DE: {}".format(str(e)))
         assert "The shape" in str(e)
 
 
+
 def test_invert_md5_py():
     """
     Test Invert python op with md5 check