File size: 4,619 Bytes
412c852
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
_base_ = [
    '../_base_/datasets/ade20k.py', '../_base_/default_runtime.py',
    '../_base_/schedules/schedule_160k.py'
]
norm_cfg = dict(type='SyncBN', requires_grad=True)
crop_size = (512, 512)
data_preprocessor = dict(
    type='SegDataPreProcessor',
    size=crop_size,
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    bgr_to_rgb=True,
    pad_val=0,
    seg_pad_val=255)
# model_cfg
num_classes = 150
model = dict(
    type='EncoderDecoder',
    data_preprocessor=data_preprocessor,
    backbone=dict(
        type='ResNet',
        depth=50,
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        dilations=(1, 1, 1, 1),
        strides=(1, 2, 2, 2),
        norm_cfg=norm_cfg,
        norm_eval=True,
        style='pytorch',
        contract_dilation=True,
        init_cfg=dict(type='Pretrained', checkpoint='torchvision://resnet50')),
    decode_head=dict(
        type='MaskFormerHead',
        in_channels=[256, 512, 1024,
                     2048],  # input channels of pixel_decoder modules
        feat_channels=256,
        in_index=[0, 1, 2, 3],
        num_classes=150,
        out_channels=256,
        num_queries=100,
        pixel_decoder=dict(
            type='mmdet.PixelDecoder',
            norm_cfg=dict(type='GN', num_groups=32),
            act_cfg=dict(type='ReLU')),
        enforce_decoder_input_project=False,
        positional_encoding=dict(  # SinePositionalEncoding
            num_feats=128, normalize=True),
        transformer_decoder=dict(  # DetrTransformerDecoder
            return_intermediate=True,
            num_layers=6,
            layer_cfg=dict(  # DetrTransformerDecoderLayer
                self_attn_cfg=dict(  # MultiheadAttention
                    embed_dims=256,
                    num_heads=8,
                    attn_drop=0.1,
                    proj_drop=0.1,
                    dropout_layer=None,
                    batch_first=True),
                cross_attn_cfg=dict(  # MultiheadAttention
                    embed_dims=256,
                    num_heads=8,
                    attn_drop=0.1,
                    proj_drop=0.1,
                    dropout_layer=None,
                    batch_first=True),
                ffn_cfg=dict(
                    embed_dims=256,
                    feedforward_channels=2048,
                    num_fcs=2,
                    act_cfg=dict(type='ReLU', inplace=True),
                    ffn_drop=0.1,
                    dropout_layer=None,
                    add_identity=True)),
            init_cfg=None),
        loss_cls=dict(
            type='mmdet.CrossEntropyLoss',
            use_sigmoid=False,
            loss_weight=1.0,
            reduction='mean',
            class_weight=[1.0] * num_classes + [0.1]),
        loss_mask=dict(
            type='mmdet.FocalLoss',
            use_sigmoid=True,
            gamma=2.0,
            alpha=0.25,
            reduction='mean',
            loss_weight=20.0),
        loss_dice=dict(
            type='mmdet.DiceLoss',
            use_sigmoid=True,
            activate=True,
            reduction='mean',
            naive_dice=True,
            eps=1.0,
            loss_weight=1.0),
        train_cfg=dict(
            assigner=dict(
                type='mmdet.HungarianAssigner',
                match_costs=[
                    dict(type='mmdet.ClassificationCost', weight=1.0),
                    dict(
                        type='mmdet.FocalLossCost',
                        weight=20.0,
                        binary_input=True),
                    dict(
                        type='mmdet.DiceCost',
                        weight=1.0,
                        pred_act=True,
                        eps=1.0)
                ]),
            sampler=dict(type='mmdet.MaskPseudoSampler'))),
    # training and testing settings
    train_cfg=dict(),
    test_cfg=dict(mode='whole'),
)
# optimizer
optimizer = dict(
    type='AdamW', lr=0.0001, betas=(0.9, 0.999), weight_decay=0.0001)
optim_wrapper = dict(
    _delete_=True,
    type='OptimWrapper',
    optimizer=optimizer,
    clip_grad=dict(max_norm=0.01, norm_type=2),
    paramwise_cfg=dict(custom_keys={
        'backbone': dict(lr_mult=0.1),
    }))
# learning policy
param_scheduler = [
    dict(
        type='PolyLR',
        eta_min=0,
        power=0.9,
        begin=0,
        end=160000,
        by_epoch=False)
]

# In MaskFormer implementation we use batch size 2 per GPU as default
train_dataloader = dict(batch_size=2, num_workers=2)
val_dataloader = dict(batch_size=1, num_workers=4)
test_dataloader = val_dataloader