Add loss evaluator (#678)

* Fix license in setup.py

* Add code for loss evaluator

* Configs support loss evaluator

* Fix a little bug

* Fix flake8

* return revised bbox to reg

* return revised bbox to reg

* revision according to comments

* fix flake8

configs/fast_rcnn_r101_fpn_1x.py

@@ -28,7 +28,13 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss',
+            use_sigmoid=False,
+            loss_weight=1.0),
+        loss_bbox=dict(
+            type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rcnn=dict(

configs/fast_rcnn_r50_caffe_c4_1x.py

@@ -36,7 +36,13 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss',
+            use_sigmoid=False,
+            loss_weight=1.0),
+        loss_bbox=dict(
+            type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rcnn=dict(

configs/fast_rcnn_r50_fpn_1x.py

@@ -28,7 +28,13 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss',
+            use_sigmoid=False,
+            loss_weight=1.0),
+        loss_bbox=dict(
+            type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rcnn=dict(

configs/faster_rcnn_ohem_r50_fpn_1x.py

@@ -23,7 +23,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -38,7 +40,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -56,7 +61,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/faster_rcnn_r101_fpn_1x.py

@@ -23,7 +23,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -38,7 +40,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -56,7 +61,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/faster_rcnn_r50_caffe_c4_1x.py

@@ -32,7 +32,9 @@ model = dict(
         anchor_strides=[16],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -46,7 +48,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -64,7 +69,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/faster_rcnn_r50_fpn_1x.py

@@ -23,7 +23,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -38,7 +40,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -56,7 +61,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/faster_rcnn_x101_32x4d_fpn_1x.py

@@ -25,7 +25,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -40,7 +42,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -58,7 +63,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,
@@ -159,7 +163,7 @@ log_config = dict(
 total_epochs = 12
 dist_params = dict(backend='nccl')
 log_level = 'INFO'
-work_dir = './work_dirs/faster_rcnn_r50_fpn_1x'
+work_dir = './work_dirs/faster_rcnn_x101_32x4d_fpn_1x'
 load_from = None
 resume_from = None
 workflow = [('train', 1)]

configs/faster_rcnn_x101_64x4d_fpn_1x.py

@@ -25,7 +25,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -40,7 +42,10 @@ model = dict(
         num_classes=81,
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
-        reg_class_agnostic=False))
+        reg_class_agnostic=False,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -58,7 +63,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,
@@ -159,7 +163,7 @@ log_config = dict(
 total_epochs = 12
 dist_params = dict(backend='nccl')
 log_level = 'INFO'
-work_dir = './work_dirs/faster_rcnn_r50_fpn_1x'
+work_dir = './work_dirs/faster_rcnn_x101_64x4d_fpn_1x'
 load_from = None
 resume_from = None
 workflow = [('train', 1)]

configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py

@@ -29,7 +29,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -48,7 +50,10 @@ model = dict(
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg))
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -66,7 +71,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py

@@ -29,7 +29,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -48,7 +50,10 @@ model = dict(
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -61,7 +66,9 @@ model = dict(
         conv_out_channels=256,
         num_classes=81,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg))
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -79,7 +86,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py

@@ -29,7 +29,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -48,7 +50,10 @@ model = dict(
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -61,7 +66,9 @@ model = dict(
         conv_out_channels=256,
         num_classes=81,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg))
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -79,7 +86,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py

@@ -31,7 +31,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -50,7 +52,10 @@ model = dict(
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -63,7 +68,9 @@ model = dict(
         conv_out_channels=256,
         num_classes=81,
         conv_cfg=conv_cfg,
-        norm_cfg=norm_cfg))
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -81,7 +88,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn/mask_rcnn_r101_fpn_gn_2x.py

@@ -27,7 +27,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -45,7 +47,10 @@ model = dict(
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -57,8 +62,9 @@ model = dict(
         in_channels=256,
         conv_out_channels=256,
         num_classes=81,
-        norm_cfg=norm_cfg))
-
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -76,7 +82,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn/mask_rcnn_r50_fpn_gn_2x.py

@@ -27,7 +27,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -45,7 +47,10 @@ model = dict(
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -57,8 +62,9 @@ model = dict(
         in_channels=256,
         conv_out_channels=256,
         num_classes=81,
-        norm_cfg=norm_cfg))
-
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -76,7 +82,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/gn/mask_rcnn_r50_fpn_gn_contrib_2x.py

@@ -27,7 +27,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -45,7 +47,10 @@ model = dict(
         target_means=[0., 0., 0., 0.],
         target_stds=[0.1, 0.1, 0.2, 0.2],
         reg_class_agnostic=False,
-        norm_cfg=norm_cfg),
+        norm_cfg=norm_cfg,
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0, loss_weight=1.0)),
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -57,8 +62,9 @@ model = dict(
         in_channels=256,
         conv_out_channels=256,
         num_classes=81,
-        norm_cfg=norm_cfg))
-
+        norm_cfg=norm_cfg,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)))
 # model training and testing settings
 train_cfg = dict(
     rpn=dict(
@@ -76,7 +82,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/htc/README.md

@@ -34,7 +34,7 @@ mmdetection
 
 ## Results and Models
 
-The results on COCO 2017val is shown in the below table. (results on test-dev are usually slightly higher than val)
+The results on COCO 2017val are shown in the below table. (results on test-dev are usually slightly higher than val)
 
 | Backbone  | Style   | Lr schd | Mem (GB) | Train time (s/iter) | Inf time (fps) | box AP | mask AP | Download |
 |:---------:|:-------:|:-------:|:--------:|:-------------------:|:--------------:|:------:|:-------:|:--------:|
@@ -52,4 +52,4 @@ We also provide a powerful HTC with DCN and multi-scale training model. No testi
 
 | Backbone         | Style   | DCN   | training scales | Lr schd | box AP | mask AP | Download |
 |:----------------:|:-------:|:-----:|:---------------:|:-------:|:------:|:-------:|:--------:|
-| X-101-64x4d-FPN  | pytorch | c3-c5 | 400~1400        | 20e     | 50.7   | 43.9    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e_20190408-0e50669c.pth) |
+| X-101-64x4d-FPN  | pytorch | c3-c5 | 400~1400        | 20e     | 50.7   | 43.9    | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e_20190408-0e50669c.pth) |
\ No newline at end of file

configs/htc/htc_dconv_c3-c5_mstrain_400_1400_x101_64x4d_fpn_20e.py

@@ -34,7 +34,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -50,7 +52,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.1, 0.1, 0.2, 0.2],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -60,7 +70,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.05, 0.05, 0.1, 0.1],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -70,7 +88,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.033, 0.033, 0.067, 0.067],
-            reg_class_agnostic=True)
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0))
     ],
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
@@ -82,7 +108,9 @@ model = dict(
         num_convs=4,
         in_channels=256,
         conv_out_channels=256,
-        num_classes=81),
+        num_classes=81,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)),
     semantic_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -115,7 +143,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/htc/htc_r101_fpn_20e.py

@@ -26,7 +26,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -42,7 +44,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.1, 0.1, 0.2, 0.2],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -52,7 +62,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.05, 0.05, 0.1, 0.1],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -62,7 +80,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.033, 0.033, 0.067, 0.067],
-            reg_class_agnostic=True)
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0))
     ],
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
@@ -74,7 +100,9 @@ model = dict(
         num_convs=4,
         in_channels=256,
         conv_out_channels=256,
-        num_classes=81),
+        num_classes=81,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)),
     semantic_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -107,7 +135,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,

configs/htc/htc_r50_fpn_1x.py

@@ -26,7 +26,9 @@ model = dict(
         anchor_strides=[4, 8, 16, 32, 64],
         target_means=[.0, .0, .0, .0],
         target_stds=[1.0, 1.0, 1.0, 1.0],
-        use_sigmoid_cls=True),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0)),
     bbox_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
@@ -42,7 +44,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.1, 0.1, 0.2, 0.2],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -52,7 +62,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.05, 0.05, 0.1, 0.1],
-            reg_class_agnostic=True),
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0)),
         dict(
             type='SharedFCBBoxHead',
             num_fcs=2,
@@ -62,7 +80,15 @@ model = dict(
             num_classes=81,
             target_means=[0., 0., 0., 0.],
             target_stds=[0.033, 0.033, 0.067, 0.067],
-            reg_class_agnostic=True)
+            reg_class_agnostic=True,
+            loss_cls=dict(
+                type='CrossEntropyLoss',
+                use_sigmoid=False,
+                loss_weight=1.0),
+            loss_bbox=dict(
+                type='SmoothL1Loss',
+                beta=1.0,
+                loss_weight=1.0))
     ],
     mask_roi_extractor=dict(
         type='SingleRoIExtractor',
@@ -74,7 +100,9 @@ model = dict(
         num_convs=4,
         in_channels=256,
         conv_out_channels=256,
-        num_classes=81),
+        num_classes=81,
+        loss_mask=dict(
+            type='CrossEntropyLoss', use_mask=True, loss_weight=1.0)),
     semantic_roi_extractor=dict(
         type='SingleRoIExtractor',
         roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
@@ -107,7 +135,6 @@ train_cfg = dict(
             add_gt_as_proposals=False),
         allowed_border=0,
         pos_weight=-1,
-        smoothl1_beta=1 / 9.0,
         debug=False),
     rpn_proposal=dict(
         nms_across_levels=False,