[MaskRCNN] refactor into generalized rcnn

examples/FasterRCNN/README.md

@@ -91,9 +91,9 @@ Mask R-CNN results contain both box and mask mAP.
  | R50-FPN  | 42.0;36.3                                                                                                                     |                                                    | 41h             | <details><summary>+Cascade</summary>`MODE_FPN=True FPN.CASCADE=True` </details>                                                                                                                                                                        |
  | R50-FPN  | 39.5;35.2                                                                                                                     | 39.5;34.4<sup>[2](#ft2)</sup>                      | 33h             | <details><summary>+ConvGNHead</summary>`MODE_FPN=True`<br/>`FPN.FRCNN_HEAD_FUNC=fastrcnn_4conv1fc_gn_head` </details>                                                                                                                                  |
  | R50-FPN  | 40.0;36.2 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R50FPN-MaskRCNN-StandardGN.npz)                         | 40.3;35.7                                          | 40h             | <details><summary>+GN</summary>`MODE_FPN=True`<br/>`FPN.NORM=GN BACKBONE.NORM=GN`<br/>`FPN.FRCNN_HEAD_FUNC=fastrcnn_4conv1fc_gn_head`<br/>`FPN.MRCNN_HEAD_FUNC=maskrcnn_up4conv_gn_head`                                                               |
- | R101-C4  | 41.4;35.2 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101C4-MaskRCNN-Standard.npz)                           |                                                    | 60h             | <details><summary>standard</summary>`BACKBONE.RESNET_NUM_BLOCK=[3,4,23,3]` </details>                                                                                                                                                                  |
- | R101-FPN | 40.4;36.6 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101FPN-MaskRCNN-Standard.npz)                          | 40.9;36.4                                          | 38h             | <details><summary>standard</summary>`MODE_FPN=True`<br/>`BACKBONE.RESNET_NUM_BLOCK=[3,4,23,3]` </details>                                                                                                                                              |
- | R101-FPN | 46.5;40.1 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101FPN-MaskRCNN-BetterParams.npz) <sup>[3](#ft3)</sup> |                                                    | 73h             | <details><summary>+++</summary>`MODE_FPN=True FPN.CASCADE=True`<br/>`BACKBONE.RESNET_NUM_BLOCK=[3,4,23,3]`<br/>`TEST.RESULT_SCORE_THRESH=1e-4`<br/>`PREPROC.TRAIN_SHORT_EDGE_SIZE=[640,800]`<br/>`TRAIN.LR_SCHEDULE=[420000,500000,540000]` </details> |
+ | R101-C4  | 41.4;35.2 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101C4-MaskRCNN-Standard.npz)                           |                                                    | 60h             | <details><summary>standard</summary>`BACKBONE.RESNET_NUM_BLOCKS=[3,4,23,3]` </details>                                                                                                                                                                  |
+ | R101-FPN | 40.4;36.6 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101FPN-MaskRCNN-Standard.npz)                          | 40.9;36.4                                          | 38h             | <details><summary>standard</summary>`MODE_FPN=True`<br/>`BACKBONE.RESNET_NUM_BLOCKS=[3,4,23,3]` </details>                                                                                                                                              |
+ | R101-FPN | 46.5;40.1 [:arrow_down:](http://models.tensorpack.com/FasterRCNN/COCO-R101FPN-MaskRCNN-BetterParams.npz) <sup>[3](#ft3)</sup> |                                                    | 73h             | <details><summary>+++</summary>`MODE_FPN=True FPN.CASCADE=True`<br/>`BACKBONE.RESNET_NUM_BLOCKS=[3,4,23,3]`<br/>`TEST.RESULT_SCORE_THRESH=1e-4`<br/>`PREPROC.TRAIN_SHORT_EDGE_SIZE=[640,800]`<br/>`TRAIN.LR_SCHEDULE=[420000,500000,540000]` </details> |
 
  <a id="ft1">1</a>: Here we comapre models that have identical training & inference cost between the two implementation. However their numbers are different due to many small implementation details.
 

examples/FasterRCNN/config.py

@@ -86,8 +86,8 @@ _C.DATA.CLASS_NAMES = []  # NUM_CLASS (NUM_CATEGORY+1) strings, to be populated
 
 # basemodel ----------------------
 _C.BACKBONE.WEIGHTS = ''   # /path/to/weights.npz
-_C.BACKBONE.RESNET_NUM_BLOCK = [3, 4, 6, 3]     # for resnet50
-# RESNET_NUM_BLOCK = [3, 4, 23, 3]    # for resnet101
+_C.BACKBONE.RESNET_NUM_BLOCKS = [3, 4, 6, 3]     # for resnet50
+# RESNET_NUM_BLOCKS = [3, 4, 23, 3]    # for resnet101
 _C.BACKBONE.FREEZE_AFFINE = False   # do not train affine parameters inside norm layers
 _C.BACKBONE.NORM = 'FreezeBN'  # options: FreezeBN, SyncBN, GN
 _C.BACKBONE.FREEZE_AT = 2  # options: 0, 1, 2

examples/FasterRCNN/model_fpn.py

@@ -160,7 +160,7 @@ def multilevel_rpn_losses(
         total_label_loss = tf.add_n(losses[::2], name='label_loss')
         total_box_loss = tf.add_n(losses[1::2], name='box_loss')
         add_moving_summary(total_label_loss, total_box_loss)
-    return total_label_loss, total_box_loss
+    return [total_label_loss, total_box_loss]
 
 
 @under_name_scope()
@@ -179,11 +179,11 @@ def generate_fpn_proposals(
     assert len(multilevel_pred_boxes) == num_lvl
     assert len(multilevel_label_logits) == num_lvl
 
-    ctx = get_current_tower_context()
+    training = get_current_tower_context().is_training
     all_boxes = []
     all_scores = []
     if cfg.FPN.PROPOSAL_MODE == 'Level':
-        fpn_nms_topk = cfg.RPN.TRAIN_PER_LEVEL_NMS_TOPK if ctx.is_training else cfg.RPN.TEST_PER_LEVEL_NMS_TOPK
+        fpn_nms_topk = cfg.RPN.TRAIN_PER_LEVEL_NMS_TOPK if training else cfg.RPN.TEST_PER_LEVEL_NMS_TOPK
         for lvl in range(num_lvl):
             with tf.name_scope('Lvl{}'.format(lvl + 2)):
                 pred_boxes_decoded = multilevel_pred_boxes[lvl]
@@ -210,8 +210,8 @@ def generate_fpn_proposals(
         all_scores = tf.concat(all_scores, axis=0)
         proposal_boxes, proposal_scores = generate_rpn_proposals(
             all_boxes, all_scores, image_shape2d,
-            cfg.RPN.TRAIN_PRE_NMS_TOPK if ctx.is_training else cfg.RPN.TEST_PRE_NMS_TOPK,
-            cfg.RPN.TRAIN_POST_NMS_TOPK if ctx.is_training else cfg.RPN.TEST_POST_NMS_TOPK)
+            cfg.RPN.TRAIN_PRE_NMS_TOPK if training else cfg.RPN.TEST_PRE_NMS_TOPK,
+            cfg.RPN.TRAIN_POST_NMS_TOPK if training else cfg.RPN.TEST_POST_NMS_TOPK)
 
     tf.sigmoid(proposal_scores, name='probs')  # for visualization
     return tf.stop_gradient(proposal_boxes, name='boxes'), \

examples/FasterRCNN/model_frcnn.py

@@ -171,7 +171,7 @@ def fastrcnn_losses(labels, label_logits, fg_boxes, fg_box_logits):
 
     add_moving_summary(label_loss, box_loss, accuracy,
                        fg_accuracy, false_negative, tf.to_float(num_fg, name='num_fg_label'))
-    return label_loss, box_loss
+    return [label_loss, box_loss]
 
 
 @under_name_scope()

examples/FasterRCNN/model_rpn.py

@@ -98,7 +98,7 @@ def rpn_losses(anchor_labels, anchor_boxes, label_logits, box_logits):
     box_loss = tf.where(tf.equal(nr_pos, 0), placeholder, box_loss, name='box_loss')
 
     add_moving_summary(label_loss, box_loss, nr_valid, nr_pos)
-    return label_loss, box_loss
+    return [label_loss, box_loss]
 
 
 @under_name_scope()