datapreprocessor, formatting

mmedit/datasets/transforms/formatting.py

@@ -182,7 +182,7 @@ def transform(self, results: dict) -> dict:
 
         data_sample.set_metainfo(metainfo=metainfo)
 
-        packed_results['data_sample'] = data_sample
+        packed_results['data_samples'] = data_sample
 
         return packed_results
 

mmedit/evaluation/metrics/matting.py

@@ -84,7 +84,7 @@ def __init__(
         super().__init__(**kwargs)
 
     def process(self, data_batch: Sequence[dict],
-                predictions: Sequence[dict]) -> None:
+                data_samples: Sequence[dict]) -> None:
         """Process one batch of data and predictions
 
         Args:
@@ -93,7 +93,7 @@ def process(self, data_batch: Sequence[dict],
             predictions (Sequence[dict]): A batch of outputs from
                 the model.
         """
-        for data, prediction in zip(data_batch, predictions):
+        for data, prediction in zip(data_batch, data_samples):
             pred_alpha, gt_alpha, _ = _fetch_data_and_check(data, prediction)
 
             # divide by 1000 to reduce the magnitude of the result

mmedit/models/data_preprocessors/data_processor.py

@@ -103,7 +103,9 @@ def forward(
             model input.
         """
 
-        inputs, batch_data_samples = self.collate_data(data)
+        # inputs, batch_data_samples = self.collate_data(data)
+        data = super().forward(data=data, training=training)
+        inputs, batch_data_samples = data['inputs'], data['data_samples']
 
         # Check if input is normalized to [0, 1]
         self.norm_input_flag = (inputs[0].max() <= 1)
@@ -113,17 +115,19 @@ def forward(
                   for _input in inputs]
 
         # Pad and stack Tensor.
-        batch_inputs, self.padded_sizes = stack_batch(inputs,
-                                                      self.pad_size_divisor,
-                                                      self.pad_args)
+        inputs, self.padded_sizes = stack_batch(inputs, self.pad_size_divisor,
+                                                self.pad_args)
 
         if training:
             for data_sample in batch_data_samples:
                 data_sample.gt_img.data = (
                     (data_sample.gt_img.data - self.outputs_mean[0]) /
                     self.outputs_std[0])
 
-        return batch_inputs, batch_data_samples
+        data['inputs'] = inputs
+        data['data_samples'] = batch_data_samples
+        # return inputs, batch_data_samples
+        return data
 
     def destructor(self, batch_tensor: torch.Tensor):
         """Destructor of data processor.

mmedit/models/data_preprocessors/mattor_preprocessor.py

@@ -135,7 +135,7 @@ def forward(self,
             # Image may of different size when testing
             assert len(data) == 1, ('only batch_size=1 '
                                     'is supported for testing.')
-
+        data = super().forward(data, training=training)
         images, trimaps, batch_data_samples = self.collate_data(data)
 
         batch_images = self._proc_inputs(images)
@@ -164,16 +164,19 @@ def forward(self,
         # N, (4/6), H, W
         batch_inputs = torch.cat((batch_images, batch_trimaps), dim=1)
 
-        return batch_inputs, batch_data_samples
+        data['inputs'] = batch_inputs
+        data['data_samples'] = batch_data_samples
+        # return batch_inputs, batch_data_samples
+        return data
 
     def collate_data(self, data: Sequence[dict]) -> Tuple[list, list, list]:
         """Collating and moving data to the target device.
 
         See base class ``BaseDataPreprocessor`` for detailed information.
         """
         inputs = [data_['inputs'] for data_ in data]
-        trimaps = [data_['data_sample'].trimap.data for data_ in data]
-        batch_data_samples = [data_['data_sample'] for data_ in data]
+        trimaps = [data_['data_samples'].trimap.data for data_ in data]
+        batch_data_samples = [data_['data_samples'] for data_ in data]
 
         # Move data from CPU to corresponding device.
         inputs = [_input.to(self.device) for _input in inputs]