res2net.py 11.6 KB
Newer Older
S
shangliang Xu 已提交
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from numbers import Integral

import paddle
import paddle.nn as nn
import paddle.nn.functional as F
from ppdet.core.workspace import register, serializable
from ..shape_spec import ShapeSpec
from .resnet import ConvNormLayer

__all__ = ['Res2Net', 'Res2NetC5']

Res2Net_cfg = {
    50: [3, 4, 6, 3],
    101: [3, 4, 23, 3],
    152: [3, 8, 36, 3],
    200: [3, 12, 48, 3]
}


class BottleNeck(nn.Layer):
    def __init__(self,
                 ch_in,
                 ch_out,
                 stride,
                 shortcut,
                 width,
                 scales=4,
                 variant='b',
                 groups=1,
                 lr=1.0,
                 norm_type='bn',
                 norm_decay=0.,
                 freeze_norm=True,
                 dcn_v2=False):
        super(BottleNeck, self).__init__()

        self.shortcut = shortcut
        self.scales = scales
        self.stride = stride
        if not shortcut:
            if variant == 'd' and stride == 2:
                self.branch1 = nn.Sequential()
                self.branch1.add_sublayer(
                    'pool',
                    nn.AvgPool2D(
                        kernel_size=2, stride=2, padding=0, ceil_mode=True))
                self.branch1.add_sublayer(
                    'conv',
                    ConvNormLayer(
                        ch_in=ch_in,
                        ch_out=ch_out,
                        filter_size=1,
                        stride=1,
                        norm_type=norm_type,
                        norm_decay=norm_decay,
                        freeze_norm=freeze_norm,
                        lr=lr))
            else:
                self.branch1 = ConvNormLayer(
                    ch_in=ch_in,
                    ch_out=ch_out,
                    filter_size=1,
                    stride=stride,
                    norm_type=norm_type,
                    norm_decay=norm_decay,
                    freeze_norm=freeze_norm,
                    lr=lr)

        self.branch2a = ConvNormLayer(
            ch_in=ch_in,
            ch_out=width * scales,
            filter_size=1,
            stride=stride if variant == 'a' else 1,
            groups=1,
            act='relu',
            norm_type=norm_type,
            norm_decay=norm_decay,
            freeze_norm=freeze_norm,
            lr=lr)

        self.branch2b = nn.LayerList([
            ConvNormLayer(
                ch_in=width,
                ch_out=width,
                filter_size=3,
                stride=1 if variant == 'a' else stride,
                groups=groups,
                act='relu',
                norm_type=norm_type,
                norm_decay=norm_decay,
                freeze_norm=freeze_norm,
                lr=lr,
                dcn_v2=dcn_v2) for _ in range(self.scales - 1)
        ])

        self.branch2c = ConvNormLayer(
            ch_in=width * scales,
            ch_out=ch_out,
            filter_size=1,
            stride=1,
            groups=1,
            norm_type=norm_type,
            norm_decay=norm_decay,
            freeze_norm=freeze_norm,
            lr=lr)

    def forward(self, inputs):

        out = self.branch2a(inputs)
        feature_split = paddle.split(out, self.scales, 1)
        out_split = []
        for i in range(self.scales - 1):
            if i == 0 or self.stride == 2:
                out_split.append(self.branch2b[i](feature_split[i]))
            else:
                out_split.append(self.branch2b[i](paddle.add(feature_split[i],
                                                             out_split[-1])))
        if self.stride == 1:
            out_split.append(feature_split[-1])
        else:
            out_split.append(F.avg_pool2d(feature_split[-1], 3, self.stride, 1))
        out = self.branch2c(paddle.concat(out_split, 1))

        if self.shortcut:
            short = inputs
        else:
            short = self.branch1(inputs)

        out = paddle.add(out, short)
        out = F.relu(out)

        return out


class Blocks(nn.Layer):
    def __init__(self,
                 ch_in,
                 ch_out,
                 count,
                 stage_num,
                 width,
                 scales=4,
                 variant='b',
                 groups=1,
                 lr=1.0,
                 norm_type='bn',
                 norm_decay=0.,
                 freeze_norm=True,
                 dcn_v2=False):
        super(Blocks, self).__init__()

        self.blocks = nn.Sequential()
        for i in range(count):
            self.blocks.add_sublayer(
                str(i),
                BottleNeck(
                    ch_in=ch_in if i == 0 else ch_out,
                    ch_out=ch_out,
                    stride=2 if i == 0 and stage_num != 2 else 1,
                    shortcut=False if i == 0 else True,
                    width=width * (2**(stage_num - 2)),
                    scales=scales,
                    variant=variant,
                    groups=groups,
                    lr=lr,
                    norm_type=norm_type,
                    norm_decay=norm_decay,
                    freeze_norm=freeze_norm,
                    dcn_v2=dcn_v2))

    def forward(self, inputs):
        return self.blocks(inputs)


@register
@serializable
class Res2Net(nn.Layer):
    """
    Res2Net, see https://arxiv.org/abs/1904.01169
    Args:
        depth (int): Res2Net depth, should be 50, 101, 152, 200.
        width (int): Res2Net width
        scales (int): Res2Net scale
        variant (str): Res2Net variant, supports 'a', 'b', 'c', 'd' currently
        lr_mult_list (list): learning rate ratio of different resnet stages(2,3,4,5),
                             lower learning rate ratio is need for pretrained model
                             got using distillation(default as [1.0, 1.0, 1.0, 1.0]).
        groups (int): The groups number of the Conv Layer.
        norm_type (str): normalization type, 'bn' or 'sync_bn'
        norm_decay (float): weight decay for normalization layer weights
        freeze_norm (bool): freeze normalization layers
        freeze_at (int): freeze the backbone at which stage
        return_idx (list): index of stages whose feature maps are returned,
                           index 0 stands for res2
        dcn_v2_stages (list): index of stages who select deformable conv v2
        num_stages (int): number of stages created

    """
    __shared__ = ['norm_type']

    def __init__(self,
                 depth=50,
                 width=26,
                 scales=4,
                 variant='b',
                 lr_mult_list=[1.0, 1.0, 1.0, 1.0],
                 groups=1,
                 norm_type='bn',
                 norm_decay=0.,
                 freeze_norm=True,
                 freeze_at=0,
                 return_idx=[0, 1, 2, 3],
                 dcn_v2_stages=[-1],
                 num_stages=4):
        super(Res2Net, self).__init__()

        self._model_type = 'Res2Net' if groups == 1 else 'Res2NeXt'

        assert depth in [50, 101, 152, 200], \
            "depth {} not in [50, 101, 152, 200]"
        assert variant in ['a', 'b', 'c', 'd'], "invalid Res2Net variant"
        assert num_stages >= 1 and num_stages <= 4

        self.depth = depth
        self.variant = variant
        self.norm_type = norm_type
        self.norm_decay = norm_decay
        self.freeze_norm = freeze_norm
        self.freeze_at = freeze_at
        if isinstance(return_idx, Integral):
            return_idx = [return_idx]
        assert max(return_idx) < num_stages, \
            'the maximum return index must smaller than num_stages, ' \
            'but received maximum return index is {} and num_stages ' \
            'is {}'.format(max(return_idx), num_stages)
        self.return_idx = return_idx
        self.num_stages = num_stages
        assert len(lr_mult_list) == 4, \
            "lr_mult_list length must be 4 but got {}".format(len(lr_mult_list))
        if isinstance(dcn_v2_stages, Integral):
            dcn_v2_stages = [dcn_v2_stages]
        assert max(dcn_v2_stages) < num_stages
        self.dcn_v2_stages = dcn_v2_stages

        block_nums = Res2Net_cfg[depth]

        # C1 stage
        if self.variant in ['c', 'd']:
            conv_def = [
                [3, 32, 3, 2, "conv1_1"],
                [32, 32, 3, 1, "conv1_2"],
                [32, 64, 3, 1, "conv1_3"],
            ]
        else:
            conv_def = [[3, 64, 7, 2, "conv1"]]
        self.res1 = nn.Sequential()
        for (c_in, c_out, k, s, _name) in conv_def:
            self.res1.add_sublayer(
                _name,
                ConvNormLayer(
                    ch_in=c_in,
                    ch_out=c_out,
                    filter_size=k,
                    stride=s,
                    groups=1,
                    act='relu',
                    norm_type=norm_type,
                    norm_decay=norm_decay,
                    freeze_norm=freeze_norm,
                    lr=1.0))

        self._in_channels = [64, 256, 512, 1024]
        self._out_channels = [256, 512, 1024, 2048]
        self._out_strides = [4, 8, 16, 32]

        # C2-C5 stages
        self.res_layers = []
        for i in range(num_stages):
            lr_mult = lr_mult_list[i]
            stage_num = i + 2
            self.res_layers.append(
                self.add_sublayer(
                    "res{}".format(stage_num),
                    Blocks(
                        self._in_channels[i],
                        self._out_channels[i],
                        count=block_nums[i],
                        stage_num=stage_num,
                        width=width,
                        scales=scales,
                        groups=groups,
                        lr=lr_mult,
                        norm_type=norm_type,
                        norm_decay=norm_decay,
                        freeze_norm=freeze_norm,
                        dcn_v2=(i in self.dcn_v2_stages))))

    @property
    def out_shape(self):
        return [
            ShapeSpec(
                channels=self._out_channels[i], stride=self._out_strides[i])
            for i in self.return_idx
        ]

    def forward(self, inputs):
        x = inputs['image']
        res1 = self.res1(x)
        x = F.max_pool2d(res1, kernel_size=3, stride=2, padding=1)
        outs = []
        for idx, stage in enumerate(self.res_layers):
            x = stage(x)
            if idx == self.freeze_at:
                x.stop_gradient = True
            if idx in self.return_idx:
                outs.append(x)
        return outs


@register
class Res2NetC5(nn.Layer):
    def __init__(self, depth=50, width=26, scales=4, variant='b'):
        super(Res2NetC5, self).__init__()
        feat_in, feat_out = [1024, 2048]
        self.res5 = Blocks(
            feat_in,
            feat_out,
            count=3,
            stage_num=5,
            width=width,
            scales=scales,
            variant=variant)
        self.feat_out = feat_out

    @property
    def out_shape(self):
        return [ShapeSpec(
            channels=self.feat_out,
            stride=32, )]

    def forward(self, roi_feat, stage=0):
        y = self.res5(roi_feat)
        return y