add dcn

xieenze · Oct 22, 2019 · ccdec9f · ccdec9f
1 parent 3111f83
commit ccdec9f
Show file tree

Hide file tree

Showing 4 changed files with 433 additions and 1 deletion.
diff --git a/README.md b/README.md
@@ -47,9 +47,15 @@ Trained models can be download in [Google Drive](https://drive.google.com/drive/
 | R-50      | caffe   | Y    | N       | 1x      |  32    | 8.9/23.9      | 29.1   | 
 | R-101     | caffe   | Y    | N       | 1x      |  32    | -             | 30.4   |
 | X-101     | pytorch | Y    | N       | 1x      |  32    | -             | 32.6   | 
+| R-50-DCN  | caffe   | Y    | N       | 1x      |  32    | -             | 32.0   | 
+| R-101-DCN | caffe   | Y    | N       | 1x      |  32    | -             | 33.5   |
+| X-101-DCN | pytorch | Y    | N       | 1x      |  32    | -             | 34.9   | 
 | R-50      | caffe   | Y    | Y       | 2x      |  32    | 8.9/23.9      | 30.5   | 
 | R-101     | caffe   | Y    | Y       | 2x      |  32    | -             | 31.9   |
-| X-101     | pytorch | Y    | Y       | 2x      |  32    | -             | 33.5   | 
+| X-101     | pytorch | Y    | Y       | 2x      |  32    | -             | 33.5   |
+| R-50-DCN  | caffe   | Y    | Y       | 2x      |  32    | -             | 33.3   | 
+| R-101-DCN | caffe   | Y    | Y       | 2x      |  32    | -             | xx.x   |
+| X-101-DCN | pytorch | Y    | Y       | 2x      |  32    | -             | xx.x   |  
 
 **Notes:**
 - The X-101 backbone is X-101-64x4d.

diff --git a/configs/polarmask/32gpu/polar_640_800_2x_r101_dcn2.py b/configs/polarmask/32gpu/polar_640_800_2x_r101_dcn2.py
@@ -0,0 +1,141 @@
+# model settings
+model = dict(
+    type='FCOS',
+    pretrained='open-mmlab://resnet101_caffe',
+    backbone=dict(
+        type='ResNet',
+        depth=101,
+        num_stages=4,
+        out_indices=(0, 1, 2, 3),
+        frozen_stages=1,
+        norm_cfg=dict(type='BN', requires_grad=False),
+        style='caffe',
+        dcn = dict(
+            modulated=False,
+            deformable_groups=1,
+            fallback_on_stride=False),
+      stage_with_dcn = (False, True, True, True)),
+    neck=dict(
+        type='FPN',
+        in_channels=[256, 512, 1024, 2048],
+        out_channels=256,
+        start_level=1,
+        add_extra_convs=True,
+        extra_convs_on_inputs=False,  # use P5
+        num_outs=5,
+        relu_before_extra_convs=True),
+    bbox_head=dict(
+        type='FCOS_Instance_Head_MIOU_MSKCTNESS',
+        num_classes=81,
+        in_channels=256,
+        stacked_convs=4,
+        feat_channels=256,
+        strides=[8, 16, 32, 64, 128],
+        use_dcn=True,
+        loss_cls=dict(
+            type='FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_bbox=dict(type='IoULoss', loss_weight=1.0),
+        loss_centerness=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)))
+# training and testing settings
+train_cfg = dict(
+    assigner=dict(
+        type='MaxIoUAssigner',
+        pos_iou_thr=0.5,
+        neg_iou_thr=0.4,
+        min_pos_iou=0,
+        ignore_iof_thr=-1),
+    allowed_border=-1,
+    pos_weight=-1,
+    debug=False)
+test_cfg = dict(
+    nms_pre=1000,
+    min_bbox_size=0,
+    score_thr=0.05,
+    nms=dict(type='nms', iou_thr=0.5),
+    max_per_img=100)
+# dataset settings
+dataset_type = 'Coco_Seg_Dataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+    mean=[102.9801, 115.9465, 122.7717], std=[1.0, 1.0, 1.0], to_rgb=False)
+data = dict(
+    imgs_per_gpu=1,
+    workers_per_gpu=4,
+    train=dict(
+        type=dataset_type,
+        ann_file=data_root + 'annotations/instances_train2017.json',
+        img_prefix=data_root + 'train2017/',
+        img_scale=[(1333, 640), (1333, 800)],
+        multiscale_mode='range',
+        img_norm_cfg=img_norm_cfg,
+        size_divisor=128,
+        flip_ratio=0.5,
+        with_mask=True,
+        with_crowd=False,
+        with_label=True,
+        resize_keep_ratio=False),
+    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=128,
+        flip_ratio=0,
+        with_mask=False,
+        with_crowd=False,
+        with_label=True,
+        resize_keep_ratio=False),
+    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=128,
+        flip_ratio=0,
+        with_mask=False,
+        with_crowd=False,
+        with_label=False,
+        resize_keep_ratio=False,
+        test_mode=True))
+# optimizer
+lr_ratio = 2
+
+optimizer = dict(
+    type='SGD',
+    lr=0.01 * lr_ratio,
+    momentum=0.9,
+    weight_decay=0.0001,
+    paramwise_options=dict(bias_lr_mult=2., bias_decay_mult=0.))
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+    policy='step',
+    warmup='linear',
+    warmup_iters=2000 ,
+    warmup_ratio=1.0 / 3 / lr_ratio,
+    step=[16+2, 22+2])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+    interval=10,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook')
+    ])
+# yapf:enable
+# runtime settings
+total_epochs = 24+2
+device_ids = range(4)
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/trash'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]
diff --git a/configs/polarmask/32gpu/polar_640_800_2x_r50_dcn2.py b/configs/polarmask/32gpu/polar_640_800_2x_r50_dcn2.py
@@ -0,0 +1,141 @@
+# model settings
+model = dict(
+    type='FCOS',
+    pretrained='open-mmlab://resnet50_caffe',
+    backbone=dict(
+        type='ResNet',
+        depth=50,
+        num_stages=4,
+        out_indices=(0, 1, 2, 3),
+        frozen_stages=1,
+        norm_cfg=dict(type='BN', requires_grad=False),
+        style='caffe',
+        dcn = dict(
+            modulated=False,
+            deformable_groups=1,
+            fallback_on_stride=False),
+      stage_with_dcn = (False, True, True, True)),
+    neck=dict(
+        type='FPN',
+        in_channels=[256, 512, 1024, 2048],
+        out_channels=256,
+        start_level=1,
+        add_extra_convs=True,
+        extra_convs_on_inputs=False,  # use P5
+        num_outs=5,
+        relu_before_extra_convs=True),
+    bbox_head=dict(
+        type='FCOS_Instance_Head_MIOU_MSKCTNESS',
+        num_classes=81,
+        in_channels=256,
+        stacked_convs=4,
+        feat_channels=256,
+        strides=[8, 16, 32, 64, 128],
+        use_dcn=True,
+        loss_cls=dict(
+            type='FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_bbox=dict(type='IoULoss', loss_weight=1.0),
+        loss_centerness=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)))
+# training and testing settings
+train_cfg = dict(
+    assigner=dict(
+        type='MaxIoUAssigner',
+        pos_iou_thr=0.5,
+        neg_iou_thr=0.4,
+        min_pos_iou=0,
+        ignore_iof_thr=-1),
+    allowed_border=-1,
+    pos_weight=-1,
+    debug=False)
+test_cfg = dict(
+    nms_pre=1000,
+    min_bbox_size=0,
+    score_thr=0.05,
+    nms=dict(type='nms', iou_thr=0.5),
+    max_per_img=100)
+# dataset settings
+dataset_type = 'Coco_Seg_Dataset'
+data_root = 'data/coco/'
+img_norm_cfg = dict(
+    mean=[102.9801, 115.9465, 122.7717], std=[1.0, 1.0, 1.0], to_rgb=False)
+data = dict(
+    imgs_per_gpu=1,
+    workers_per_gpu=5,
+    train=dict(
+        type=dataset_type,
+        ann_file=data_root + 'annotations/instances_train2017.json',
+        img_prefix=data_root + 'train2017/',
+        img_scale=[(1333, 640), (1333, 800)],
+        multiscale_mode='range',
+        img_norm_cfg=img_norm_cfg,
+        size_divisor=128,
+        flip_ratio=0.5,
+        with_mask=True,
+        with_crowd=False,
+        with_label=True,
+        resize_keep_ratio=False),
+    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=128,
+        flip_ratio=0,
+        with_mask=False,
+        with_crowd=False,
+        with_label=True,
+        resize_keep_ratio=False),
+    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=128,
+        flip_ratio=0,
+        with_mask=False,
+        with_crowd=False,
+        with_label=False,
+        resize_keep_ratio=False,
+        test_mode=True))
+# optimizer
+lr_ratio = 2
+
+optimizer = dict(
+    type='SGD',
+    lr=0.01 * lr_ratio,
+    momentum=0.9,
+    weight_decay=0.0001,
+    paramwise_options=dict(bias_lr_mult=2., bias_decay_mult=0.))
+optimizer_config = dict(grad_clip=dict(max_norm=35, norm_type=2))
+# learning policy
+lr_config = dict(
+    policy='step',
+    warmup='linear',
+    warmup_iters=2000 ,
+    warmup_ratio=1.0 / 3 / lr_ratio,
+    step=[16+2, 22+2])
+checkpoint_config = dict(interval=1)
+# yapf:disable
+log_config = dict(
+    interval=10,
+    hooks=[
+        dict(type='TextLoggerHook'),
+        # dict(type='TensorboardLoggerHook')
+    ])
+# yapf:enable
+# runtime settings
+total_epochs = 24+2
+device_ids = range(4)
+dist_params = dict(backend='nccl')
+log_level = 'INFO'
+work_dir = './work_dirs/trash'
+load_from = None
+resume_from = None
+workflow = [('train', 1)]