fix problem of s2anet while deploying (#4624)

* fix problem of s2anet while deploying

* correct problem of s2anet TestReader

* add dota default category

configs/datasets/dota.yml

@@ -17,4 +17,3 @@ EvalDataset:
 TestDataset:
   !ImageFolder
     anno_path: trainval_split/s2anet_trainval_paddle_coco.json
-    dataset_dir: dataset/DOTA_1024_s2anet/

configs/dota/_base_/s2anet_reader.yml

@@ -29,8 +29,6 @@ EvalReader:
 
 
 TestReader:
-  inputs_def:
-    image_shape: [3, 1024, 1024]
   sample_transforms:
   - Decode: {}
   - Resize: {interp: 2, target_size: [1024, 1024], keep_ratio: True}

configs/dota/s2anet_alignconv_2x_dota.yml

@@ -24,4 +24,3 @@ S2ANetHead:
   reg_loss_weight: [1.0, 1.0, 1.0, 1.0, 1.1]
   cls_loss_weight: [1.1, 1.05]
   reg_loss_type: 'l1'
-  is_training: True

configs/dota/s2anet_conv_2x_dota.yml

@@ -28,4 +28,3 @@ S2ANetHead:
   use_sigmoid_cls: True
   reg_loss_weight: [1.0, 1.0, 1.0, 1.0, 1.1]
   cls_loss_weight: [1.1, 1.05]
-  is_training: True

ppdet/data/source/category.py

@@ -39,7 +39,8 @@ def get_categories(metric_type, anno_file=None, arch=None):
     if arch == 'keypoint_arch':
         return (None, {'id': 'keypoint'})
 
-    if metric_type.lower() == 'coco' or metric_type.lower() == 'rbox' or metric_type.lower() == 'snipercoco':
+    if metric_type.lower() == 'coco' or metric_type.lower(
+    ) == 'rbox' or metric_type.lower() == 'snipercoco':
         if anno_file and os.path.isfile(anno_file):
             # lazy import pycocotools here
             from pycocotools.coco import COCO
@@ -53,6 +54,9 @@ def get_categories(metric_type, anno_file=None, arch=None):
 
         # anno file not exist, load default categories of COCO17
         else:
+            if metric_type.lower() == 'rbox':
+                return _dota_category()
+
             return _coco17_category()
 
     elif metric_type.lower() == 'voc':
@@ -294,6 +298,34 @@ def _coco17_category():
     return clsid2catid, catid2name
 
 
+def _dota_category():
+    """
+    Get class id to category id map and category id
+    to category name map of dota dataset
+    """
+    catid2name = {
+        0: 'background',
+        1: 'plane',
+        2: 'baseball-diamond',
+        3: 'bridge',
+        4: 'ground-track-field',
+        5: 'small-vehicle',
+        6: 'large-vehicle',
+        7: 'ship',
+        8: 'tennis-court',
+        9: 'basketball-court',
+        10: 'storage-tank',
+        11: 'soccer-ball-field',
+        12: 'roundabout',
+        13: 'harbor',
+        14: 'swimming-pool',
+        15: 'helicopter'
+    }
+    catid2name.pop(0)
+    clsid2catid = {i: i + 1 for i in range(len(catid2name))}
+    return clsid2catid, catid2name
+
+
 def _vocall_category():
     """
     Get class id to category id map and category id

ppdet/modeling/heads/s2anet_head.py

@@ -233,8 +233,7 @@ class S2ANetHead(nn.Layer):
                  anchor_assign=RBoxAssigner().__dict__,
                  reg_loss_weight=[1.0, 1.0, 1.0, 1.0, 1.1],
                  cls_loss_weight=[1.1, 1.05],
-                 reg_loss_type='l1',
-                 is_training=True):
+                 reg_loss_type='l1'):
         super(S2ANetHead, self).__init__()
         self.stacked_convs = stacked_convs
         self.feat_in = feat_in
@@ -260,8 +259,6 @@ class S2ANetHead(nn.Layer):
         self.alpha = 1.0
         self.beta = 1.0
         self.reg_loss_type = reg_loss_type
-        self.is_training = is_training
-
         self.s2anet_head_out = None
 
         # anchor
@@ -451,12 +448,10 @@ class S2ANetHead(nn.Layer):
             init_anchors = self.rect2rbox(init_anchors)
             self.base_anchors_list.append(init_anchors)
 
-            if self.is_training:
+            if self.training:
                 refine_anchor = self.bbox_decode(fam_reg.detach(), init_anchors)
             else:
-                fam_reg1 = fam_reg.clone()
-                fam_reg1.stop_gradient = True
-                refine_anchor = self.bbox_decode(fam_reg1, init_anchors)
+                refine_anchor = self.bbox_decode(fam_reg, init_anchors)
 
             self.refine_anchor_list.append(refine_anchor)