From b5a952c51110f2ba6f9e5ba8cbad1342393a0abb Mon Sep 17 00:00:00 2001
From: Kai Chen <chenkaidev@gmail.com>
Date: Tue, 21 May 2019 11:04:26 +0800
Subject: [PATCH] Add training from scratch with GN (#658)
* add training from scratch with GN
* rename and format
* change resnet initialization
* update scratch results
---
configs/scratch/README.md | 22 ++
.../scratch_faster_rcnn_r50_fpn_gn_6x.py | 173 ++++++++++++++++
.../scratch_mask_rcnn_r50_fpn_gn_6x.py | 188 ++++++++++++++++++
3 files changed, 383 insertions(+)
create mode 100644 configs/scratch/README.md
create mode 100644 configs/scratch/scratch_faster_rcnn_r50_fpn_gn_6x.py
create mode 100644 configs/scratch/scratch_mask_rcnn_r50_fpn_gn_6x.py
diff --git a/configs/scratch/README.md b/configs/scratch/README.md
new file mode 100644
index 0000000..7ae532b
--- /dev/null
+++ b/configs/scratch/README.md
@@ -0,0 +1,22 @@
+# Rethinking ImageNet Pre-training
+
+## Introduction
+
+```
+@article{he2018rethinking,
+ title={Rethinking imagenet pre-training},
+ author={He, Kaiming and Girshick, Ross and Doll{\'a}r, Piotr},
+ journal={arXiv preprint arXiv:1811.08883},
+ year={2018}
+}
+```
+
+## Results and Models
+
+| Model | Backbone | Style | Lr schd | box AP | mask AP | Download |
+|:------------:|:---------:|:-------:|:-------:|:------:|:-------:|:--------:|
+| Faster R-CNN | R-50-FPN | pytorch | 6x | 40.1 | - | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/scratch/scratch_faster_rcnn_r50_fpn_gn_6x-20190515-ff554978.pth) |
+| Mask R-CNN | R-50-FPN | pytorch | 6x | 41.0 | 37.4 | [model](https://s3.ap-northeast-2.amazonaws.com/open-mmlab/mmdetection/models/scratch/scratch_mask_rcnn_r50_fpn_gn_6x_20190515-96743f5e.pth) |
+
+Note:
+- The above models are trained with 16 GPUs.
\ No newline at end of file
diff --git a/configs/scratch/scratch_faster_rcnn_r50_fpn_gn_6x.py b/configs/scratch/scratch_faster_rcnn_r50_fpn_gn_6x.py
new file mode 100644
index 0000000..d4da529
--- /dev/null
+++ b/configs/scratch/scratch_faster_rcnn_r50_fpn_gn_6x.py
@@ -0,0 +1,173 @@
+# model settings
+norm_cfg = dict(type='GN', num_groups=32, requires_grad=True)
+model = dict(
+ type='FasterRCNN',
+ pretrained=None,
+ backbone=dict(
+ type='ResNet',
+ depth=50,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=-1,
+ style='pytorch',
+ zero_init_residual=False,
+ norm_cfg=norm_cfg),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ norm_cfg=norm_cfg),
+ 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,
+ norm_cfg=norm_cfg))
+# 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),
+ rpn_proposal=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ 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=1000,
+ nms_post=1000,
+ max_num=1000,
+ 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,
+ paramwise_options=dict(norm_decay_mult=0))
+optimizer_config = dict(grad_clip=None)
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=0.1,
+ step=[65, 71])
+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 = 73
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/scratch_faster_rcnn_r50_fpn_gn_6x'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/configs/scratch/scratch_mask_rcnn_r50_fpn_gn_6x.py b/configs/scratch/scratch_mask_rcnn_r50_fpn_gn_6x.py
new file mode 100644
index 0000000..63d21c8
--- /dev/null
+++ b/configs/scratch/scratch_mask_rcnn_r50_fpn_gn_6x.py
@@ -0,0 +1,188 @@
+# model settings
+norm_cfg = dict(type='GN', num_groups=32, requires_grad=True)
+model = dict(
+ type='MaskRCNN',
+ pretrained=None,
+ backbone=dict(
+ type='ResNet',
+ depth=50,
+ num_stages=4,
+ out_indices=(0, 1, 2, 3),
+ frozen_stages=-1,
+ style='pytorch',
+ zero_init_residual=False,
+ norm_cfg=norm_cfg),
+ neck=dict(
+ type='FPN',
+ in_channels=[256, 512, 1024, 2048],
+ out_channels=256,
+ num_outs=5,
+ norm_cfg=norm_cfg),
+ 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,
+ norm_cfg=norm_cfg),
+ 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,
+ norm_cfg=norm_cfg))
+# 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),
+ rpn_proposal=dict(
+ nms_across_levels=False,
+ nms_pre=2000,
+ nms_post=2000,
+ max_num=2000,
+ nms_thr=0.7,
+ min_bbox_size=0),
+ 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=1000,
+ nms_post=1000,
+ max_num=1000,
+ 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,
+ paramwise_options=dict(norm_decay_mult=0))
+optimizer_config = dict(grad_clip=None)
+# learning policy
+lr_config = dict(
+ policy='step',
+ warmup='linear',
+ warmup_iters=500,
+ warmup_ratio=0.1,
+ step=[65, 71])
+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 = 73
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/scratch_mask_rcnn_r50_fpn_gn_6x'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
--
GitLab