Reorder data feed (#2592)

PaddleCV/object_detection/ppdet/data/data_feed.py

@@ -81,7 +81,7 @@ def create_reader(feed, max_iter=0):
             'TYPE': type(feed.dataset).__source__
         }
     }
-    
+
     if len(getattr(feed.dataset, 'images', [])) > 0:
         data_config[mode]['IMAGES'] = feed.dataset.images
 
@@ -448,43 +448,33 @@ class FasterRCNNTrainFeed(DataFeed):
         self.mode = 'TRAIN'
 
 
-# XXX currently use two presets, in the future, these should be combined into a
-# single `RCNNTrainFeed`. Mask (and keypoint) should be processed
-# automatically if `gt_mask` (or `gt_keypoints`) is in the required fields
 @register
-class MaskRCNNTrainFeed(DataFeed):
+class FasterRCNNEvalFeed(DataFeed):
     __doc__ = DataFeed.__doc__
 
     def __init__(self,
-                 dataset=CocoDataSet().__dict__,
-                 fields=[
-                     'image', 'im_info', 'im_id', 'gt_box', 'gt_label',
-                     'is_crowd', 'gt_mask'
-                 ],
+                 dataset=CocoDataSet(COCO_VAL_ANNOTATION,
+                                     COCO_VAL_IMAGE_DIR).__dict__,
+                 fields=['image', 'im_info', 'im_id', 'im_shape'],
                  image_shape=[3, 1333, 800],
                  sample_transforms=[
                      DecodeImage(to_rgb=True),
-                     RandomFlipImage(prob=0.5, is_mask_flip=True),
                      NormalizeImage(mean=[0.485, 0.456, 0.406],
                                     std=[0.229, 0.224, 0.225],
                                     is_scale=True,
                                     is_channel_first=False),
-                     ResizeImage(target_size=800,
-                                 max_size=1333,
-                                 interp=1,
-                                 use_cv2=True),
-                     Permute(to_bgr=False, channel_first=True)
+                     ResizeImage(target_size=800, max_size=1333, interp=1),
+                     Permute(to_bgr=False)
                  ],
                  batch_transforms=[PadBatch()],
                  batch_size=1,
-                 shuffle=True,
+                 shuffle=False,
                  samples=-1,
                  drop_last=False,
                  num_workers=2,
-                 use_process=False,
-                 use_padded_im_info=False):
-        sample_transforms.append(ArrangeRCNN(is_mask=True))
-        super(MaskRCNNTrainFeed, self).__init__(
+                 use_padded_im_info=True):
+        sample_transforms.append(ArrangeTestRCNN())
+        super(FasterRCNNEvalFeed, self).__init__(
             dataset,
             fields,
             image_shape,
@@ -495,17 +485,17 @@ class MaskRCNNTrainFeed(DataFeed):
             samples=samples,
             drop_last=drop_last,
             num_workers=num_workers,
-            use_process=use_process)
-        self.mode = 'TRAIN'
+            use_padded_im_info=use_padded_im_info)
+        self.mode = 'VAL'
 
 
 @register
-class FasterRCNNEvalFeed(DataFeed):
+class FasterRCNNTestFeed(DataFeed):
     __doc__ = DataFeed.__doc__
 
     def __init__(self,
-                 dataset=CocoDataSet(COCO_VAL_ANNOTATION,
-                                     COCO_VAL_IMAGE_DIR).__dict__,
+                 dataset=SimpleDataSet(COCO_VAL_ANNOTATION,
+                                       COCO_VAL_IMAGE_DIR).__dict__,
                  fields=['image', 'im_info', 'im_id', 'im_shape'],
                  image_shape=[3, 1333, 800],
                  sample_transforms=[
@@ -514,7 +504,6 @@ class FasterRCNNEvalFeed(DataFeed):
                                     std=[0.229, 0.224, 0.225],
                                     is_scale=True,
                                     is_channel_first=False),
-                     ResizeImage(target_size=800, max_size=1333, interp=1),
                      Permute(to_bgr=False)
                  ],
                  batch_transforms=[PadBatch()],
@@ -525,7 +514,9 @@ class FasterRCNNEvalFeed(DataFeed):
                  num_workers=2,
                  use_padded_im_info=True):
         sample_transforms.append(ArrangeTestRCNN())
-        super(FasterRCNNEvalFeed, self).__init__(
+        if isinstance(dataset, dict):
+            dataset = SimpleDataSet(**dataset)
+        super(FasterRCNNTestFeed, self).__init__(
             dataset,
             fields,
             image_shape,
@@ -537,37 +528,46 @@ class FasterRCNNEvalFeed(DataFeed):
             drop_last=drop_last,
             num_workers=num_workers,
             use_padded_im_info=use_padded_im_info)
-        self.mode = 'VAL'
+        self.mode = 'TEST'
 
 
+# XXX currently use two presets, in the future, these should be combined into a
+# single `RCNNTrainFeed`. Mask (and keypoint) should be processed
+# automatically if `gt_mask` (or `gt_keypoints`) is in the required fields
 @register
-class FasterRCNNTestFeed(DataFeed):
+class MaskRCNNTrainFeed(DataFeed):
     __doc__ = DataFeed.__doc__
 
     def __init__(self,
-                 dataset=SimpleDataSet(COCO_VAL_ANNOTATION,
-                                       COCO_VAL_IMAGE_DIR).__dict__,
-                 fields=['image', 'im_info', 'im_id', 'im_shape'],
+                 dataset=CocoDataSet().__dict__,
+                 fields=[
+                     'image', 'im_info', 'im_id', 'gt_box', 'gt_label',
+                     'is_crowd', 'gt_mask'
+                 ],
                  image_shape=[3, 1333, 800],
                  sample_transforms=[
                      DecodeImage(to_rgb=True),
+                     RandomFlipImage(prob=0.5, is_mask_flip=True),
                      NormalizeImage(mean=[0.485, 0.456, 0.406],
                                     std=[0.229, 0.224, 0.225],
                                     is_scale=True,
                                     is_channel_first=False),
-                     Permute(to_bgr=False)
+                     ResizeImage(target_size=800,
+                                 max_size=1333,
+                                 interp=1,
+                                 use_cv2=True),
+                     Permute(to_bgr=False, channel_first=True)
                  ],
                  batch_transforms=[PadBatch()],
                  batch_size=1,
-                 shuffle=False,
+                 shuffle=True,
                  samples=-1,
                  drop_last=False,
                  num_workers=2,
-                 use_padded_im_info=True):
-        sample_transforms.append(ArrangeTestRCNN())
-        if isinstance(dataset, dict):
-            dataset = SimpleDataSet(**dataset)
-        super(FasterRCNNTestFeed, self).__init__(
+                 use_process=False,
+                 use_padded_im_info=False):
+        sample_transforms.append(ArrangeRCNN(is_mask=True))
+        super(MaskRCNNTrainFeed, self).__init__(
             dataset,
             fields,
             image_shape,
@@ -578,8 +578,8 @@ class FasterRCNNTestFeed(DataFeed):
             samples=samples,
             drop_last=drop_last,
             num_workers=num_workers,
-            use_padded_im_info=use_padded_im_info)
-        self.mode = 'TEST'
+            use_process=use_process)
+        self.mode = 'TRAIN'
 
 
 @register