Fix fpn issue in paddle 1.7.x

hub_module/modules/image/object_detection/faster_rcnn_resnet50_coco2017/module.py

@@ -89,7 +89,7 @@ class FasterRCNNResNet50(hub.Module):
         with fluid.program_guard(context_prog, startup_program):
             with fluid.unique_name.guard():
                 image = fluid.layers.data(
-                    name='image', shape=[3, 800, 1333], dtype='float32')
+                    name='image', shape=[-1, 3, -1, -1], dtype='float32')
                 # backbone
                 backbone = ResNet(
                     norm_type='affine_channel',
@@ -201,9 +201,13 @@ class FasterRCNNResNet50(hub.Module):
 
                 place = fluid.CPUPlace()
                 exe = fluid.Executor(place)
+                exe.run(startup_program)
                 if pretrained:
 
                     def _if_exist(var):
+                        if num_classes != 81:
+                            if 'bbox_pred' in var.name or 'cls_score' in var.name:
+                                return False
                         return os.path.exists(
                             os.path.join(self.default_pretrained_model_path,
                                          var.name))
@@ -212,8 +216,6 @@ class FasterRCNNResNet50(hub.Module):
                         exe,
                         self.default_pretrained_model_path,
                         predicate=_if_exist)
-                else:
-                    exe.run(startup_program)
                 return inputs, outputs, context_prog
 
     def rpn_head(self):

hub_module/modules/image/object_detection/faster_rcnn_resnet50_fpn_coco2017/module.py

@@ -90,7 +90,7 @@ class FasterRCNNResNet50RPN(hub.Module):
         with fluid.program_guard(context_prog, startup_program):
             with fluid.unique_name.guard():
                 image = fluid.layers.data(
-                    name='image', shape=[3, 800, 1333], dtype='float32')
+                    name='image', shape=[-1, 3, -1, -1], dtype='float32')
                 # backbone
                 backbone = ResNet(
                     norm_type='affine_channel',
@@ -202,9 +202,13 @@ class FasterRCNNResNet50RPN(hub.Module):
 
                 place = fluid.CPUPlace()
                 exe = fluid.Executor(place)
+                exe.run(startup_program)
                 if pretrained:
 
                     def _if_exist(var):
+                        if num_classes != 81:
+                            if 'bbox_pred' in var.name or 'cls_score' in var.name:
+                                return False
                         return os.path.exists(
                             os.path.join(self.default_pretrained_model_path,
                                          var.name))
@@ -213,8 +217,6 @@ class FasterRCNNResNet50RPN(hub.Module):
                         exe,
                         self.default_pretrained_model_path,
                         predicate=_if_exist)
-                else:
-                    exe.run(startup_program)
                 return inputs, outputs, context_prog
 
     def rpn_head(self):