From 6fe5ccde85d170f52e3fdaa50a9a288e93443fa5 Mon Sep 17 00:00:00 2001
From: Kai Chen <chenkaidev@gmail.com>
Date: Fri, 26 Apr 2019 21:36:55 -0700
Subject: [PATCH] Add support for weight standardization (#521)
* add support for weight standardization
* add ws support for htc heads
* add a config file for 20-23-24e lr schedule
* update readme of weight standardization
---
README.md | 1 +
configs/gn+ws/README.md | 54 ++++++
configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py | 165 ++++++++++++++++
.../mask_rcnn_r50_fpn_gn_ws_20_23_24e.py | 180 +++++++++++++++++
configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py | 180 +++++++++++++++++
.../mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py | 182 ++++++++++++++++++
mmdet/models/backbones/resnet.py | 71 ++++---
mmdet/models/backbones/resnext.py | 24 ++-
mmdet/models/bbox_heads/convfc_bbox_head.py | 3 +
mmdet/models/mask_heads/fcn_mask_head.py | 3 +
.../models/mask_heads/fused_semantic_head.py | 5 +
mmdet/models/mask_heads/htc_mask_head.py | 1 +
mmdet/models/necks/fpn.py | 3 +
mmdet/models/utils/__init__.py | 8 +-
mmdet/models/utils/conv_module.py | 41 +++-
mmdet/models/utils/conv_ws.py | 46 +++++
16 files changed, 935 insertions(+), 32 deletions(-)
create mode 100644 configs/gn+ws/README.md
create mode 100644 configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py
create mode 100644 configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py
create mode 100644 configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py
create mode 100644 configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py
create mode 100644 mmdet/models/utils/conv_ws.py
diff --git a/README.md b/README.md
index a6b40e4..bffd59a 100644
--- a/README.md
+++ b/README.md
@@ -96,6 +96,7 @@ Results and models are available in the [Model zoo](MODEL_ZOO.md).
Other features
- [x] DCNv2
- [x] Group Normalization
+- [x] Weight Standardization
- [x] OHEM
- [x] Soft-NMS
diff --git a/configs/gn+ws/README.md b/configs/gn+ws/README.md
new file mode 100644
index 0000000..511f22c
--- /dev/null
+++ b/configs/gn+ws/README.md
@@ -0,0 +1,54 @@
+# Weight Standardization
+
+## Introduction
+
+```
+@article{weightstandardization,
+ author = {Siyuan Qiao and Huiyu Wang and Chenxi Liu and Wei Shen and Alan Yuille},
+ title = {Weight Standardization},
+ journal = {arXiv preprint arXiv:1903.10520},
+ year = {2019},
+}
+```
+
+## Results and Models
+
+Faster R-CNN
+
+| Backbone | Style | Normalization | Lr schd | box AP | mask AP | Download |
+|:---------:|:-------:|:-------------:|:-------:|:------:|:-------:|:--------:|
+| R-50-FPN | pytorch | GN | 1x | 37.8 | - | - |
+| R-50-FPN | pytorch | GN+WS | 1x | 38.9 | - | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/faster_rcnn_r50_fpn_gn_ws_1x_20190418-935d00b6.pth) |
+| R-101-FPN | pytorch | GN | 1x | 39.8 | - | - |
+| R-101-FPN | pytorch | GN+WS | 1x | 41.4 | - | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/faster_rcnn_r101_fpn_gn_ws_1x_20190419-728705ec.pth) |
+| X-50-32x4d-FPN | pytorch | GN | 1x | 36.5 | - | - |
+| X-50-32x4d-FPN | pytorch | GN+WS | 1x | 39.9 | - | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/faster_rcnn_x50_32x4d_fpn_gn_ws_1x_20190419-4e61072b.pth) |
+| X-101-32x4d-FPN | pytorch | GN | 1x | 33.2 | - | - |
+| X-101-32x4d-FPN | pytorch | GN+WS | 1x | 41.8 | - | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/faster_rcnn_x101_32x4d_fpn_gn_ws_1x_20190419-c78e5583.pth) |
+
+Mask R-CNN
+
+| Backbone | Style | Normalization | Lr schd | box AP | mask AP | Download |
+|:---------:|:-------:|:-------------:|:-------:|:------:|:-------:|:--------:|
+| R-50-FPN | pytorch | GN | 2x | 39.9 | 36.0 | - |
+| R-50-FPN | pytorch | GN+WS | 2x | 40.3 | 36.2 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r50_fpn_gn_ws_2x_20190419-9ec97bbb.pth) |
+| R-101-FPN | pytorch | GN | 2x | 41.6 | 37.3 | - |
+| R-101-FPN | pytorch | GN+WS | 2x | 42.0 | 37.3 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r101_fpn_gn_ws_2x_20190419-bc7399a6.pth) |
+| X-50-32x4d-FPN | pytorch | GN | 2x | 39.2 | 35.5 | - |
+| X-50-32x4d-FPN | pytorch | GN+WS | 2x | 40.7 | 36.7 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_x50_32x4d_fpn_gn_ws_2x_20190419-2110205e.pth) |
+| X-101-32x4d-FPN | pytorch | GN | 2x | 36.4 | 33.1 | - |
+| X-101-32x4d-FPN | pytorch | GN+WS | 2x | 42.1 | 37.7 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x_20190419-7777b15f.pth) |
+| R-50-FPN | pytorch | GN | 20-23-24e | 40.6 | 36.6 | - |
+| R-50-FPN | pytorch | GN+WS | 20-23-24e | 41.1 | 37.0 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e_20190425-1d9e499e.pth) |
+| R-101-FPN | pytorch | GN | 20-23-24e | 42.3 | 38.1 | - |
+| R-101-FPN | pytorch | GN+WS | 20-23-24e | 43.0 | 38.4 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_r101_fpn_gn_ws_20_23_24e_20190425-66cb3792.pth) |
+| X-50-32x4d-FPN | pytorch | GN | 20-23-24e | 39.6 | 35.9 | - |
+| X-50-32x4d-FPN | pytorch | GN+WS | 20-23-24e | 41.9 | 37.7 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_x50_32x4d_fpn_gn_ws_20_23_24e_20190425-d01e2200.pth) |
+| X-101-32x4d-FPN | pytorch | GN | 20-23-24e | 36.6 | 33.4 | - |
+| X-101-32x4d-FPN | pytorch | GN+WS | 20-23-24e | 43.4 | 38.7 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/ws/mask_rcnn_x101_32x4d_fpn_gn_ws_20_23_24e_20190425-1ff3e5b2.pth) |
+
+Note:
+
+- GN+WS requires about 5% more memory than GN, and it is only 5% slower than GN.
+- In the paper, a 20-23-24e lr schedule is used instead of 2x.
+- The X-50-GN and X-101-GN pretrained models are also shared by the authors.
\ No newline at end of file
diff --git a/configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py b/configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py
new file mode 100644
index 0000000..d494043
--- /dev/null
+++ b/configs/gn+ws/faster_rcnn_r50_fpn_gn_ws_1x.py
@@ -0,0 +1,165 @@
+# model settings
+conv_cfg = dict(type='ConvWS')
+normalize = dict(type='GN', num_groups=32, frozen=False)
+model = dict(
+ type='FasterRCNN',
+ pretrained='open-mmlab://jhu/resnet50_gn_ws',
+ backbone=dict(
+ type='ResNet',
+ depth=50,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=1,
+ style='pytorch',
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ rpn_head=dict(
+ type='RPNHead',
+ in_channels=256,
+ feat_channels=256,
+ anchor_scales=[8],
+ anchor_ratios=[0.5, 1.0, 2.0],
+ 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),
+ bbox_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ bbox_head=dict(
+ type='ConvFCBBoxHead',
+ num_shared_convs=4,
+ num_shared_fcs=1,
+ in_channels=256,
+ conv_out_channels=256,
+ fc_out_channels=1024,
+ roi_feat_size=7,
+ num_classes=81,
+ target_means=[0., 0., 0., 0.],
+ target_stds=[0.1, 0.1, 0.2, 0.2],
+ reg_class_agnostic=False,
+ conv_cfg=conv_cfg,
+ normalize=normalize))
+# model training and testing settings
+train_cfg = dict(
+ rpn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.7,
+ neg_iou_thr=0.3,
+ min_pos_iou=0.3,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=256,
+ pos_fraction=0.5,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=False),
+ allowed_border=0,
+ pos_weight=-1,
+ smoothl1_beta=1 / 9.0,
+ debug=False),
+ rcnn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.5,
+ neg_iou_thr=0.5,
+ min_pos_iou=0.5,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=512,
+ pos_fraction=0.25,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=True),
+ pos_weight=-1,
+ debug=False))
+test_cfg = dict(
+ rpn=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ rcnn=dict(
+ score_thr=0.05,
+ nms=dict(type='nms', iou_thr=0.5),
+ max_per_img=100))
+# dataset settings
+dataset_type = 'CocoDataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+ mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+data = dict(
+ imgs_per_gpu=2,
+ workers_per_gpu=2,
+ train=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_train2017.json',
+ img_prefix=data_root + 'train2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0.5,
+ with_mask=False,
+ with_crowd=True,
+ with_label=True),
+ val=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=False,
+ with_crowd=True,
+ with_label=True),
+ test=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=False,
+ with_label=False,
+ test_mode=True))
+# optimizer
+optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=1.0 / 3,
+ step=[8, 11])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+ interval=50,
+ hooks=[
+ dict(type='TextLoggerHook'),
+ # dict(type='TensorboardLoggerHook')
+ ])
+# yapf:enable
+# runtime settings
+total_epochs = 12
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/faster_rcnn_r50_fpn_gn_ws_1x'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py b/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py
new file mode 100644
index 0000000..2d98767
--- /dev/null
+++ b/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_20_23_24e.py
@@ -0,0 +1,180 @@
+# model settings
+conv_cfg = dict(type='ConvWS')
+normalize = dict(type='GN', num_groups=32, frozen=False)
+model = dict(
+ type='MaskRCNN',
+ pretrained='open-mmlab://jhu/resnet50_gn_ws',
+ backbone=dict(
+ type='ResNet',
+ depth=50,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=1,
+ style='pytorch',
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ rpn_head=dict(
+ type='RPNHead',
+ in_channels=256,
+ feat_channels=256,
+ anchor_scales=[8],
+ anchor_ratios=[0.5, 1.0, 2.0],
+ 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),
+ bbox_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ bbox_head=dict(
+ type='ConvFCBBoxHead',
+ num_shared_convs=4,
+ num_shared_fcs=1,
+ in_channels=256,
+ conv_out_channels=256,
+ fc_out_channels=1024,
+ roi_feat_size=7,
+ num_classes=81,
+ target_means=[0., 0., 0., 0.],
+ target_stds=[0.1, 0.1, 0.2, 0.2],
+ reg_class_agnostic=False,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ mask_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ mask_head=dict(
+ type='FCNMaskHead',
+ num_convs=4,
+ in_channels=256,
+ conv_out_channels=256,
+ num_classes=81,
+ conv_cfg=conv_cfg,
+ normalize=normalize))
+# model training and testing settings
+train_cfg = dict(
+ rpn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.7,
+ neg_iou_thr=0.3,
+ min_pos_iou=0.3,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=256,
+ pos_fraction=0.5,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=False),
+ allowed_border=0,
+ pos_weight=-1,
+ smoothl1_beta=1 / 9.0,
+ debug=False),
+ rcnn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.5,
+ neg_iou_thr=0.5,
+ min_pos_iou=0.5,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=512,
+ pos_fraction=0.25,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=True),
+ mask_size=28,
+ pos_weight=-1,
+ debug=False))
+test_cfg = dict(
+ rpn=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ rcnn=dict(
+ score_thr=0.05,
+ nms=dict(type='nms', iou_thr=0.5),
+ max_per_img=100,
+ mask_thr_binary=0.5))
+# dataset settings
+dataset_type = 'CocoDataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+ mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+data = dict(
+ imgs_per_gpu=2,
+ workers_per_gpu=2,
+ train=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_train2017.json',
+ img_prefix=data_root + 'train2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0.5,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ val=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ test=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=False,
+ with_label=False,
+ test_mode=True))
+# optimizer
+optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=1.0 / 3,
+ step=[20, 23])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+ interval=50,
+ hooks=[
+ dict(type='TextLoggerHook'),
+ # dict(type='TensorboardLoggerHook')
+ ])
+# yapf:enable
+# runtime settings
+total_epochs = 24
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/mask_rcnn_r50_fpn_gn_ws_20_23_24e'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py b/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py
new file mode 100644
index 0000000..c28c6ed
--- /dev/null
+++ b/configs/gn+ws/mask_rcnn_r50_fpn_gn_ws_2x.py
@@ -0,0 +1,180 @@
+# model settings
+conv_cfg = dict(type='ConvWS')
+normalize = dict(type='GN', num_groups=32, frozen=False)
+model = dict(
+ type='MaskRCNN',
+ pretrained='open-mmlab://jhu/resnet50_gn_ws',
+ backbone=dict(
+ type='ResNet',
+ depth=50,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=1,
+ style='pytorch',
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ rpn_head=dict(
+ type='RPNHead',
+ in_channels=256,
+ feat_channels=256,
+ anchor_scales=[8],
+ anchor_ratios=[0.5, 1.0, 2.0],
+ 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),
+ bbox_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ bbox_head=dict(
+ type='ConvFCBBoxHead',
+ num_shared_convs=4,
+ num_shared_fcs=1,
+ in_channels=256,
+ conv_out_channels=256,
+ fc_out_channels=1024,
+ roi_feat_size=7,
+ num_classes=81,
+ target_means=[0., 0., 0., 0.],
+ target_stds=[0.1, 0.1, 0.2, 0.2],
+ reg_class_agnostic=False,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ mask_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ mask_head=dict(
+ type='FCNMaskHead',
+ num_convs=4,
+ in_channels=256,
+ conv_out_channels=256,
+ num_classes=81,
+ conv_cfg=conv_cfg,
+ normalize=normalize))
+# model training and testing settings
+train_cfg = dict(
+ rpn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.7,
+ neg_iou_thr=0.3,
+ min_pos_iou=0.3,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=256,
+ pos_fraction=0.5,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=False),
+ allowed_border=0,
+ pos_weight=-1,
+ smoothl1_beta=1 / 9.0,
+ debug=False),
+ rcnn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.5,
+ neg_iou_thr=0.5,
+ min_pos_iou=0.5,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=512,
+ pos_fraction=0.25,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=True),
+ mask_size=28,
+ pos_weight=-1,
+ debug=False))
+test_cfg = dict(
+ rpn=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ rcnn=dict(
+ score_thr=0.05,
+ nms=dict(type='nms', iou_thr=0.5),
+ max_per_img=100,
+ mask_thr_binary=0.5))
+# dataset settings
+dataset_type = 'CocoDataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+ mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+data = dict(
+ imgs_per_gpu=2,
+ workers_per_gpu=2,
+ train=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_train2017.json',
+ img_prefix=data_root + 'train2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0.5,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ val=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ test=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=False,
+ with_label=False,
+ test_mode=True))
+# optimizer
+optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=1.0 / 3,
+ step=[16, 22])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+ interval=50,
+ hooks=[
+ dict(type='TextLoggerHook'),
+ # dict(type='TensorboardLoggerHook')
+ ])
+# yapf:enable
+# runtime settings
+total_epochs = 24
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/mask_rcnn_r50_fpn_gn_ws_2x'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py b/configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py
new file mode 100644
index 0000000..8fdeaa0
--- /dev/null
+++ b/configs/gn+ws/mask_rcnn_x101_32x4d_fpn_gn_ws_2x.py
@@ -0,0 +1,182 @@
+# model settings
+conv_cfg = dict(type='ConvWS')
+normalize = dict(type='GN', num_groups=32, frozen=False)
+model = dict(
+ type='MaskRCNN',
+ pretrained='open-mmlab://jhu/resnext101_32x4d_gn_ws',
+ backbone=dict(
+ type='ResNeXt',
+ depth=101,
+ groups=32,
+ base_width=4,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=1,
+ style='pytorch',
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ rpn_head=dict(
+ type='RPNHead',
+ in_channels=256,
+ feat_channels=256,
+ anchor_scales=[8],
+ anchor_ratios=[0.5, 1.0, 2.0],
+ 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),
+ bbox_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=7, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ bbox_head=dict(
+ type='ConvFCBBoxHead',
+ num_shared_convs=4,
+ num_shared_fcs=1,
+ in_channels=256,
+ conv_out_channels=256,
+ fc_out_channels=1024,
+ roi_feat_size=7,
+ num_classes=81,
+ target_means=[0., 0., 0., 0.],
+ target_stds=[0.1, 0.1, 0.2, 0.2],
+ reg_class_agnostic=False,
+ conv_cfg=conv_cfg,
+ normalize=normalize),
+ mask_roi_extractor=dict(
+ type='SingleRoIExtractor',
+ roi_layer=dict(type='RoIAlign', out_size=14, sample_num=2),
+ out_channels=256,
+ featmap_strides=[4, 8, 16, 32]),
+ mask_head=dict(
+ type='FCNMaskHead',
+ num_convs=4,
+ in_channels=256,
+ conv_out_channels=256,
+ num_classes=81,
+ conv_cfg=conv_cfg,
+ normalize=normalize))
+# model training and testing settings
+train_cfg = dict(
+ rpn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.7,
+ neg_iou_thr=0.3,
+ min_pos_iou=0.3,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=256,
+ pos_fraction=0.5,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=False),
+ allowed_border=0,
+ pos_weight=-1,
+ smoothl1_beta=1 / 9.0,
+ debug=False),
+ rcnn=dict(
+ assigner=dict(
+ type='MaxIoUAssigner',
+ pos_iou_thr=0.5,
+ neg_iou_thr=0.5,
+ min_pos_iou=0.5,
+ ignore_iof_thr=-1),
+ sampler=dict(
+ type='RandomSampler',
+ num=512,
+ pos_fraction=0.25,
+ neg_pos_ub=-1,
+ add_gt_as_proposals=True),
+ mask_size=28,
+ pos_weight=-1,
+ debug=False))
+test_cfg = dict(
+ rpn=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ rcnn=dict(
+ score_thr=0.05,
+ nms=dict(type='nms', iou_thr=0.5),
+ max_per_img=100,
+ mask_thr_binary=0.5))
+# dataset settings
+dataset_type = 'CocoDataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+ mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+data = dict(
+ imgs_per_gpu=2,
+ workers_per_gpu=2,
+ train=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_train2017.json',
+ img_prefix=data_root + 'train2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0.5,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ val=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=True,
+ with_crowd=True,
+ with_label=True),
+ test=dict(
+ type=dataset_type,
+ ann_file=data_root + 'annotations/instances_val2017.json',
+ img_prefix=data_root + 'val2017/',
+ img_scale=(1333, 800),
+ img_norm_cfg=img_norm_cfg,
+ size_divisor=32,
+ flip_ratio=0,
+ with_mask=False,
+ with_label=False,
+ test_mode=True))
+# optimizer
+optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0001)
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=1.0 / 3,
+ step=[16, 22])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+ interval=50,
+ hooks=[
+ dict(type='TextLoggerHook'),
+ # dict(type='TensorboardLoggerHook')
+ ])
+# yapf:enable
+# runtime settings
+total_epochs = 24
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/mask_rcnn_x101_32x4d_fpn_gn_ws_2x'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/mmdet/models/backbones/resnet.py b/mmdet/models/backbones/resnet.py
index 1965d6e..ba7e5fe 100644
--- a/mmdet/models/backbones/resnet.py
+++ b/mmdet/models/backbones/resnet.py
@@ -8,19 +8,7 @@ from mmcv.runner import load_checkpoint
from mmdet.ops import DeformConv, ModulatedDeformConv
from ..registry import BACKBONES
-from ..utils import build_norm_layer
-
-
-def conv3x3(in_planes, out_planes, stride=1, dilation=1):
- "3x3 convolution with padding"
- return nn.Conv2d(
- in_planes,
- out_planes,
- kernel_size=3,
- stride=stride,
- padding=dilation,
- dilation=dilation,
- bias=False)
+from ..utils import build_conv_layer, build_norm_layer
class BasicBlock(nn.Module):
@@ -34,6 +22,7 @@ class BasicBlock(nn.Module):
downsample=None,
style='pytorch',
with_cp=False,
+ conv_cfg=None,
normalize=dict(type='BN'),
dcn=None):
super(BasicBlock, self).__init__()
@@ -42,9 +31,25 @@ class BasicBlock(nn.Module):
self.norm1_name, norm1 = build_norm_layer(normalize, planes, postfix=1)
self.norm2_name, norm2 = build_norm_layer(normalize, planes, postfix=2)
- self.conv1 = conv3x3(inplanes, planes, stride, dilation)
+ self.conv1 = build_conv_layer(
+ conv_cfg,
+ inplanes,
+ planes,
+ 3,
+ stride=stride,
+ padding=dilation,
+ dilation=dilation,
+ bias=False)
self.add_module(self.norm1_name, norm1)
- self.conv2 = conv3x3(planes, planes)
+ self.conv2 = build_conv_layer(
+ conv_cfg,
+ planes,
+ planes,
+ 3,
+ stride=stride,
+ padding=dilation,
+ dilation=dilation,
+ bias=False)
self.add_module(self.norm2_name, norm2)
self.relu = nn.ReLU(inplace=True)
@@ -91,6 +96,7 @@ class Bottleneck(nn.Module):
downsample=None,
style='pytorch',
with_cp=False,
+ conv_cfg=None,
normalize=dict(type='BN'),
dcn=None):
"""Bottleneck block for ResNet.
@@ -102,6 +108,7 @@ class Bottleneck(nn.Module):
assert dcn is None or isinstance(dcn, dict)
self.inplanes = inplanes
self.planes = planes
+ self.conv_cfg = conv_cfg
self.normalize = normalize
self.dcn = dcn
self.with_dcn = dcn is not None
@@ -117,7 +124,8 @@ class Bottleneck(nn.Module):
self.norm3_name, norm3 = build_norm_layer(
normalize, planes * self.expansion, postfix=3)
- self.conv1 = nn.Conv2d(
+ self.conv1 = build_conv_layer(
+ conv_cfg,
inplanes,
planes,
kernel_size=1,
@@ -130,7 +138,8 @@ class Bottleneck(nn.Module):
fallback_on_stride = dcn.get('fallback_on_stride', False)
self.with_modulated_dcn = dcn.get('modulated', False)
if not self.with_dcn or fallback_on_stride:
- self.conv2 = nn.Conv2d(
+ self.conv2 = build_conv_layer(
+ conv_cfg,
planes,
planes,
kernel_size=3,
@@ -139,6 +148,7 @@ class Bottleneck(nn.Module):
dilation=dilation,
bias=False)
else:
+ assert conv_cfg is None, 'conv_cfg must be None for DCN'
deformable_groups = dcn.get('deformable_groups', 1)
if not self.with_modulated_dcn:
conv_op = DeformConv
@@ -163,8 +173,12 @@ class Bottleneck(nn.Module):
deformable_groups=deformable_groups,
bias=False)
self.add_module(self.norm2_name, norm2)
- self.conv3 = nn.Conv2d(
- planes, planes * self.expansion, kernel_size=1, bias=False)
+ self.conv3 = build_conv_layer(
+ conv_cfg,
+ planes,
+ planes * self.expansion,
+ kernel_size=1,
+ bias=False)
self.add_module(self.norm3_name, norm3)
self.relu = nn.ReLU(inplace=True)
@@ -236,12 +250,14 @@ def make_res_layer(block,
dilation=1,
style='pytorch',
with_cp=False,
+ conv_cfg=None,
normalize=dict(type='BN'),
dcn=None):
downsample = None
if stride != 1 or inplanes != planes * block.expansion:
downsample = nn.Sequential(
- nn.Conv2d(
+ build_conv_layer(
+ conv_cfg,
inplanes,
planes * block.expansion,
kernel_size=1,
@@ -260,6 +276,7 @@ def make_res_layer(block,
downsample,
style=style,
with_cp=with_cp,
+ conv_cfg=conv_cfg,
normalize=normalize,
dcn=dcn))
inplanes = planes * block.expansion
@@ -272,6 +289,7 @@ def make_res_layer(block,
dilation,
style=style,
with_cp=with_cp,
+ conv_cfg=conv_cfg,
normalize=normalize,
dcn=dcn))
@@ -319,6 +337,7 @@ class ResNet(nn.Module):
out_indices=(0, 1, 2, 3),
style='pytorch',
frozen_stages=-1,
+ conv_cfg=None,
normalize=dict(type='BN', frozen=False),
norm_eval=True,
dcn=None,
@@ -338,6 +357,7 @@ class ResNet(nn.Module):
assert max(out_indices) < num_stages
self.style = style
self.frozen_stages = frozen_stages
+ self.conv_cfg = conv_cfg
self.normalize = normalize
self.with_cp = with_cp
self.norm_eval = norm_eval
@@ -367,6 +387,7 @@ class ResNet(nn.Module):
dilation=dilation,
style=self.style,
with_cp=with_cp,
+ conv_cfg=conv_cfg,
normalize=normalize,
dcn=dcn)
self.inplanes = planes * self.block.expansion
@@ -384,8 +405,14 @@ class ResNet(nn.Module):
return getattr(self, self.norm1_name)
def _make_stem_layer(self):
- self.conv1 = nn.Conv2d(
- 3, 64, kernel_size=7, stride=2, padding=3, bias=False)
+ self.conv1 = build_conv_layer(
+ self.conv_cfg,
+ 3,
+ 64,
+ kernel_size=7,
+ stride=2,
+ padding=3,
+ bias=False)
self.norm1_name, norm1 = build_norm_layer(
self.normalize, 64, postfix=1)
self.add_module(self.norm1_name, norm1)
diff --git a/mmdet/models/backbones/resnext.py b/mmdet/models/backbones/resnext.py
index 0e83d07..3cdb1cb 100644
--- a/mmdet/models/backbones/resnext.py
+++ b/mmdet/models/backbones/resnext.py
@@ -6,7 +6,7 @@ from mmdet.ops import DeformConv, ModulatedDeformConv
from .resnet import Bottleneck as _Bottleneck
from .resnet import ResNet
from ..registry import BACKBONES
-from ..utils import build_norm_layer
+from ..utils import build_conv_layer, build_norm_layer
class Bottleneck(_Bottleneck):
@@ -30,7 +30,8 @@ class Bottleneck(_Bottleneck):
self.norm3_name, norm3 = build_norm_layer(
self.normalize, self.planes * self.expansion, postfix=3)
- self.conv1 = nn.Conv2d(
+ self.conv1 = build_conv_layer(
+ self.conv_cfg,
self.inplanes,
width,
kernel_size=1,
@@ -43,7 +44,8 @@ class Bottleneck(_Bottleneck):
fallback_on_stride = self.dcn.get('fallback_on_stride', False)
self.with_modulated_dcn = self.dcn.get('modulated', False)
if not self.with_dcn or fallback_on_stride:
- self.conv2 = nn.Conv2d(
+ self.conv2 = build_conv_layer(
+ self.conv_cfg,
width,
width,
kernel_size=3,
@@ -53,6 +55,7 @@ class Bottleneck(_Bottleneck):
groups=groups,
bias=False)
else:
+ assert self.conv_cfg is None, 'conv_cfg must be None for DCN'
groups = self.dcn.get('groups', 1)
deformable_groups = self.dcn.get('deformable_groups', 1)
if not self.with_modulated_dcn:
@@ -79,8 +82,12 @@ class Bottleneck(_Bottleneck):
deformable_groups=deformable_groups,
bias=False)
self.add_module(self.norm2_name, norm2)
- self.conv3 = nn.Conv2d(
- width, self.planes * self.expansion, kernel_size=1, bias=False)
+ self.conv3 = build_conv_layer(
+ self.conv_cfg,
+ width,
+ self.planes * self.expansion,
+ kernel_size=1,
+ bias=False)
self.add_module(self.norm3_name, norm3)
@@ -94,12 +101,14 @@ def make_res_layer(block,
base_width=4,
style='pytorch',
with_cp=False,
+ conv_cfg=None,
normalize=dict(type='BN'),
dcn=None):
downsample = None
if stride != 1 or inplanes != planes * block.expansion:
downsample = nn.Sequential(
- nn.Conv2d(
+ build_conv_layer(
+ conv_cfg,
inplanes,
planes * block.expansion,
kernel_size=1,
@@ -120,6 +129,7 @@ def make_res_layer(block,
base_width=base_width,
style=style,
with_cp=with_cp,
+ conv_cfg=conv_cfg,
normalize=normalize,
dcn=dcn))
inplanes = planes * block.expansion
@@ -134,6 +144,7 @@ def make_res_layer(block,
base_width=base_width,
style=style,
with_cp=with_cp,
+ conv_cfg=conv_cfg,
normalize=normalize,
dcn=dcn))
@@ -196,6 +207,7 @@ class ResNeXt(ResNet):
base_width=self.base_width,
style=self.style,
with_cp=self.with_cp,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
dcn=dcn)
self.inplanes = planes * self.block.expansion
diff --git a/mmdet/models/bbox_heads/convfc_bbox_head.py b/mmdet/models/bbox_heads/convfc_bbox_head.py
index c424aff..af2a5e3 100644
--- a/mmdet/models/bbox_heads/convfc_bbox_head.py
+++ b/mmdet/models/bbox_heads/convfc_bbox_head.py
@@ -24,6 +24,7 @@ class ConvFCBBoxHead(BBoxHead):
num_reg_fcs=0,
conv_out_channels=256,
fc_out_channels=1024,
+ conv_cfg=None,
normalize=None,
*args,
**kwargs):
@@ -44,6 +45,7 @@ class ConvFCBBoxHead(BBoxHead):
self.num_reg_fcs = num_reg_fcs
self.conv_out_channels = conv_out_channels
self.fc_out_channels = fc_out_channels
+ self.conv_cfg = conv_cfg
self.normalize = normalize
self.with_bias = normalize is None
@@ -101,6 +103,7 @@ class ConvFCBBoxHead(BBoxHead):
self.conv_out_channels,
3,
padding=1,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
bias=self.with_bias))
last_layer_dim = self.conv_out_channels
diff --git a/mmdet/models/mask_heads/fcn_mask_head.py b/mmdet/models/mask_heads/fcn_mask_head.py
index e1889f0..614497a 100644
--- a/mmdet/models/mask_heads/fcn_mask_head.py
+++ b/mmdet/models/mask_heads/fcn_mask_head.py
@@ -22,6 +22,7 @@ class FCNMaskHead(nn.Module):
upsample_ratio=2,
num_classes=81,
class_agnostic=False,
+ conv_cfg=None,
normalize=None):
super(FCNMaskHead, self).__init__()
if upsample_method not in [None, 'deconv', 'nearest', 'bilinear']:
@@ -37,6 +38,7 @@ class FCNMaskHead(nn.Module):
self.upsample_ratio = upsample_ratio
self.num_classes = num_classes
self.class_agnostic = class_agnostic
+ self.conv_cfg = conv_cfg
self.normalize = normalize
self.with_bias = normalize is None
@@ -51,6 +53,7 @@ class FCNMaskHead(nn.Module):
self.conv_out_channels,
self.conv_kernel_size,
padding=padding,
+ conv_cfg=conv_cfg,
normalize=normalize,
bias=self.with_bias))
upsample_in_channels = (self.conv_out_channels
diff --git a/mmdet/models/mask_heads/fused_semantic_head.py b/mmdet/models/mask_heads/fused_semantic_head.py
index 8795f66..f24adf3 100644
--- a/mmdet/models/mask_heads/fused_semantic_head.py
+++ b/mmdet/models/mask_heads/fused_semantic_head.py
@@ -30,6 +30,7 @@ class FusedSemanticHead(nn.Module):
num_classes=183,
ignore_label=255,
loss_weight=0.2,
+ conv_cfg=None,
normalize=None):
super(FusedSemanticHead, self).__init__()
self.num_ins = num_ins
@@ -40,6 +41,7 @@ class FusedSemanticHead(nn.Module):
self.num_classes = num_classes
self.ignore_label = ignore_label
self.loss_weight = loss_weight
+ self.conv_cfg = conv_cfg
self.normalize = normalize
self.with_bias = normalize is None
@@ -50,6 +52,7 @@ class FusedSemanticHead(nn.Module):
self.in_channels,
self.in_channels,
1,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
bias=self.with_bias,
inplace=False))
@@ -63,12 +66,14 @@ class FusedSemanticHead(nn.Module):
conv_out_channels,
3,
padding=1,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
bias=self.with_bias))
self.conv_embedding = ConvModule(
conv_out_channels,
conv_out_channels,
1,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
bias=self.with_bias)
self.conv_logits = nn.Conv2d(conv_out_channels, self.num_classes, 1)
diff --git a/mmdet/models/mask_heads/htc_mask_head.py b/mmdet/models/mask_heads/htc_mask_head.py
index 6f1ccf0..21f3130 100644
--- a/mmdet/models/mask_heads/htc_mask_head.py
+++ b/mmdet/models/mask_heads/htc_mask_head.py
@@ -12,6 +12,7 @@ class HTCMaskHead(FCNMaskHead):
self.conv_out_channels,
self.conv_out_channels,
1,
+ conv_cfg=self.conv_cfg,
normalize=self.normalize,
bias=self.with_bias)
diff --git a/mmdet/models/necks/fpn.py b/mmdet/models/necks/fpn.py
index 90c0f23..3e49fc4 100644
--- a/mmdet/models/necks/fpn.py
+++ b/mmdet/models/necks/fpn.py
@@ -17,6 +17,7 @@ class FPN(nn.Module):
end_level=-1,
add_extra_convs=False,
extra_convs_on_inputs=True,
+ conv_cfg=None,
normalize=None,
activation=None):
super(FPN, self).__init__()
@@ -49,6 +50,7 @@ class FPN(nn.Module):
in_channels[i],
out_channels,
1,
+ conv_cfg=conv_cfg,
normalize=normalize,
bias=self.with_bias,
activation=self.activation,
@@ -58,6 +60,7 @@ class FPN(nn.Module):
out_channels,
3,
padding=1,
+ conv_cfg=conv_cfg,
normalize=normalize,
bias=self.with_bias,
activation=self.activation,
diff --git a/mmdet/models/utils/__init__.py b/mmdet/models/utils/__init__.py
index 194eb9c..c517b07 100644
--- a/mmdet/models/utils/__init__.py
+++ b/mmdet/models/utils/__init__.py
@@ -1,9 +1,11 @@
-from .conv_module import ConvModule
+from .conv_ws import conv_ws_2d, ConvWS2d
+from .conv_module import build_conv_layer, ConvModule
from .norm import build_norm_layer
from .weight_init import (xavier_init, normal_init, uniform_init, kaiming_init,
bias_init_with_prob)
__all__ = [
- 'ConvModule', 'build_norm_layer', 'xavier_init', 'normal_init',
- 'uniform_init', 'kaiming_init', 'bias_init_with_prob'
+ 'conv_ws_2d', 'ConvWS2d', 'build_conv_layer', 'ConvModule',
+ 'build_norm_layer', 'xavier_init', 'normal_init', 'uniform_init',
+ 'kaiming_init', 'bias_init_with_prob'
]
diff --git a/mmdet/models/utils/conv_module.py b/mmdet/models/utils/conv_module.py
index b5651ec..b3bf9c7 100644
--- a/mmdet/models/utils/conv_module.py
+++ b/mmdet/models/utils/conv_module.py
@@ -3,8 +3,43 @@ import warnings
import torch.nn as nn
from mmcv.cnn import kaiming_init, constant_init
+from .conv_ws import ConvWS2d
from .norm import build_norm_layer
+conv_cfg = {
+ 'Conv': nn.Conv2d,
+ 'ConvWS': ConvWS2d,
+ # TODO: octave conv
+}
+
+
+def build_conv_layer(cfg, *args, **kwargs):
+ """ Build convolution layer
+
+ Args:
+ cfg (None or dict): cfg should contain:
+ type (str): identify conv layer type.
+ layer args: args needed to instantiate a conv layer.
+
+ Returns:
+ layer (nn.Module): created conv layer
+ """
+ if cfg is None:
+ cfg_ = dict(type='Conv')
+ else:
+ assert isinstance(cfg, dict) and 'type' in cfg
+ cfg_ = cfg.copy()
+
+ layer_type = cfg_.pop('type')
+ if layer_type not in conv_cfg:
+ raise KeyError('Unrecognized norm type {}'.format(layer_type))
+ else:
+ conv_layer = conv_cfg[layer_type]
+
+ layer = conv_layer(*args, **kwargs, **cfg_)
+
+ return layer
+
class ConvModule(nn.Module):
@@ -17,11 +52,14 @@ class ConvModule(nn.Module):
dilation=1,
groups=1,
bias=True,
+ conv_cfg=None,
normalize=None,
activation='relu',
inplace=True,
activate_last=True):
super(ConvModule, self).__init__()
+ assert conv_cfg is None or isinstance(conv_cfg, dict)
+ assert normalize is None or isinstance(normalize, dict)
self.with_norm = normalize is not None
self.with_activatation = activation is not None
self.with_bias = bias
@@ -31,7 +69,8 @@ class ConvModule(nn.Module):
if self.with_norm and self.with_bias:
warnings.warn('ConvModule has norm and bias at the same time')
- self.conv = nn.Conv2d(
+ self.conv = build_conv_layer(
+ conv_cfg,
in_channels,
out_channels,
kernel_size,
diff --git a/mmdet/models/utils/conv_ws.py b/mmdet/models/utils/conv_ws.py
new file mode 100644
index 0000000..5ccd735
--- /dev/null
+++ b/mmdet/models/utils/conv_ws.py
@@ -0,0 +1,46 @@
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+def conv_ws_2d(input,
+ weight,
+ bias=None,
+ stride=1,
+ padding=0,
+ dilation=1,
+ groups=1,
+ eps=1e-5):
+ c_in = weight.size(0)
+ weight_flat = weight.view(c_in, -1)
+ mean = weight_flat.mean(dim=1, keepdim=True).view(c_in, 1, 1, 1)
+ std = weight_flat.std(dim=1, keepdim=True).view(c_in, 1, 1, 1)
+ weight = (weight - mean) / (std + eps)
+ return F.conv2d(input, weight, bias, stride, padding, dilation, groups)
+
+
+class ConvWS2d(nn.Conv2d):
+
+ def __init__(self,
+ in_channels,
+ out_channels,
+ kernel_size,
+ stride=1,
+ padding=0,
+ dilation=1,
+ groups=1,
+ bias=True,
+ eps=1e-5):
+ super(ConvWS2d, self).__init__(
+ in_channels,
+ out_channels,
+ kernel_size,
+ stride=stride,
+ padding=padding,
+ dilation=dilation,
+ groups=groups,
+ bias=bias)
+ self.eps = eps
+
+ def forward(self, x):
+ return conv_ws_2d(x, self.weight, self.bias, self.stride, self.padding,
+ self.dilation, self.groups, self.eps)
--
GitLab