se_resnext.py

# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# 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 __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import numpy as np
import paddle
from paddle import ParamAttr
import paddle.nn as nn
import paddle.nn.functional as F
from paddle.nn import Conv2D, BatchNorm, Linear, Dropout
from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D
from paddle.nn.initializer import Uniform

import math

__all__ = ["SE_ResNeXt50_32x4d", "SE_ResNeXt101_32x4d", "SE_ResNeXt152_64x4d"]


class ConvBNLayer(nn.Layer):
    def __init__(self,
                 num_channels,
                 num_filters,
                 filter_size,
                 stride=1,
                 groups=1,
                 act=None,
                 name=None,
                 data_format='NCHW'):
        super(ConvBNLayer, self).__init__()

        self._conv = Conv2D(
            in_channels=num_channels,
            out_channels=num_filters,
            kernel_size=filter_size,
            stride=stride,
            padding=(filter_size - 1) // 2,
            groups=groups,
            weight_attr=ParamAttr(name=name + "_weights"),
            bias_attr=False,
            data_format=data_format)
        bn_name = name + '_bn'
        self._batch_norm = BatchNorm(
            num_filters,
            act=act,
            param_attr=ParamAttr(name=bn_name + '_scale'),
            bias_attr=ParamAttr(bn_name + '_offset'),
            moving_mean_name=bn_name + '_mean',
            moving_variance_name=bn_name + '_variance',
            data_layout=data_format)

    def forward(self, inputs):
        y = self._conv(inputs)
        y = self._batch_norm(y)
        return y


class BottleneckBlock(nn.Layer):
    def __init__(self,
                 num_channels,
                 num_filters,
                 stride,
                 cardinality,
                 reduction_ratio,
                 shortcut=True,
                 if_first=False,
                 name=None,
                 data_format="NCHW"):
        super(BottleneckBlock, self).__init__()

        self.conv0 = ConvBNLayer(
            num_channels=num_channels,
            num_filters=num_filters,
            filter_size=1,
            act='relu',
            name='conv' + name + '_x1',
            data_format=data_format)
        self.conv1 = ConvBNLayer(
            num_channels=num_filters,
            num_filters=num_filters,
            filter_size=3,
            groups=cardinality,
            stride=stride,
            act='relu',
            name='conv' + name + '_x2',
            data_format=data_format)
        self.conv2 = ConvBNLayer(
            num_channels=num_filters,
            num_filters=num_filters * 2 if cardinality == 32 else num_filters,
            filter_size=1,
            act=None,
            name='conv' + name + '_x3',
            data_format=data_format)
        self.scale = SELayer(
            num_channels=num_filters * 2 if cardinality == 32 else num_filters,
            num_filters=num_filters * 2 if cardinality == 32 else num_filters,
            reduction_ratio=reduction_ratio,
            name='fc' + name,
            data_format=data_format)

        if not shortcut:
            self.short = ConvBNLayer(
                num_channels=num_channels,
                num_filters=num_filters * 2
                if cardinality == 32 else num_filters,
                filter_size=1,
                stride=stride,
                name='conv' + name + '_prj',
                data_format=data_format)

        self.shortcut = shortcut

    def forward(self, inputs):
        y = self.conv0(inputs)
        conv1 = self.conv1(y)
        conv2 = self.conv2(conv1)
        scale = self.scale(conv2)

        if self.shortcut:
            short = inputs
        else:
            short = self.short(inputs)
        y = paddle.add(x=short, y=scale)
        y = F.relu(y)
        return y


class SELayer(nn.Layer):
    def __init__(self, num_channels, num_filters, reduction_ratio, name=None, data_format="NCHW"):
        super(SELayer, self).__init__()

        self.data_format = data_format
        self.pool2d_gap = AdaptiveAvgPool2D(1, data_format=self.data_format)

        self._num_channels = num_channels

        med_ch = int(num_channels / reduction_ratio)
        stdv = 1.0 / math.sqrt(num_channels * 1.0)
        self.squeeze = Linear(
            num_channels,
            med_ch,
            weight_attr=ParamAttr(
                initializer=Uniform(-stdv, stdv), name=name + "_sqz_weights"),
            bias_attr=ParamAttr(name=name + '_sqz_offset'))
        self.relu = nn.ReLU()
        stdv = 1.0 / math.sqrt(med_ch * 1.0)
        self.excitation = Linear(
            med_ch,
            num_filters,
            weight_attr=ParamAttr(
                initializer=Uniform(-stdv, stdv), name=name + "_exc_weights"),
            bias_attr=ParamAttr(name=name + '_exc_offset'))
        self.sigmoid = nn.Sigmoid()

    def forward(self, input):
        pool = self.pool2d_gap(input)
        if self.data_format == "NHWC":
            pool = paddle.squeeze(pool, axis=[1, 2])
        else:
            pool = paddle.squeeze(pool, axis=[2, 3])
        squeeze = self.squeeze(pool)
        squeeze = self.relu(squeeze)
        excitation = self.excitation(squeeze)
        excitation = self.sigmoid(excitation)
        if self.data_format == "NHWC":
            excitation = paddle.unsqueeze(excitation, axis=[1, 2])
        else:
            excitation = paddle.unsqueeze(excitation, axis=[2, 3])
        out = input * excitation
        return out


class ResNeXt(nn.Layer):
    def __init__(self, layers=50, class_dim=1000, cardinality=32, input_image_channel=3, data_format="NCHW"):
        super(ResNeXt, self).__init__()

        self.layers = layers
        self.cardinality = cardinality
        self.reduction_ratio = 16
        self.data_format = data_format
        self.input_image_channel = input_image_channel

        supported_layers = [50, 101, 152]
        assert layers in supported_layers, \
            "supported layers are {} but input layer is {}".format(
                supported_layers, layers)
        supported_cardinality = [32, 64]
        assert cardinality in supported_cardinality, \
            "supported cardinality is {} but input cardinality is {}" \
            .format(supported_cardinality, cardinality)
        if layers == 50:
            depth = [3, 4, 6, 3]
        elif layers == 101:
            depth = [3, 4, 23, 3]
        elif layers == 152:
            depth = [3, 8, 36, 3]
        num_channels = [64, 256, 512, 1024]
        num_filters = [128, 256, 512,
                       1024] if cardinality == 32 else [256, 512, 1024, 2048]
        if layers < 152:
            self.conv = ConvBNLayer(
                num_channels=self.input_image_channel,
                num_filters=64,
                filter_size=7,
                stride=2,
                act='relu',
                name="conv1",
                data_format=self.data_format)
        else:
            self.conv1_1 = ConvBNLayer(
                num_channels=self.input_image_channel,
                num_filters=64,
                filter_size=3,
                stride=2,
                act='relu',
                name="conv1",
                data_format=self.data_format)
            self.conv1_2 = ConvBNLayer(
                num_channels=64,
                num_filters=64,
                filter_size=3,
                stride=1,
                act='relu',
                name="conv2",
                data_format=self.data_format)
            self.conv1_3 = ConvBNLayer(
                num_channels=64,
                num_filters=128,
                filter_size=3,
                stride=1,
                act='relu',
                name="conv3",
                data_format=self.data_format)

        self.pool2d_max = MaxPool2D(kernel_size=3, stride=2, padding=1, data_format=self.data_format)

        self.block_list = []
        n = 1 if layers == 50 or layers == 101 else 3
        for block in range(len(depth)):
            n += 1
            shortcut = False
            for i in range(depth[block]):
                bottleneck_block = self.add_sublayer(
                    'bb_%d_%d' % (block, i),
                    BottleneckBlock(
                        num_channels=num_channels[block] if i == 0 else
                        num_filters[block] * int(64 // self.cardinality),
                        num_filters=num_filters[block],
                        stride=2 if i == 0 and block != 0 else 1,
                        cardinality=self.cardinality,
                        reduction_ratio=self.reduction_ratio,
                        shortcut=shortcut,
                        if_first=block == 0,
                        name=str(n) + '_' + str(i + 1),
                        data_format=self.data_format))
                self.block_list.append(bottleneck_block)
                shortcut = True

        self.pool2d_avg = AdaptiveAvgPool2D(1, data_format=self.data_format)

        self.pool2d_avg_channels = num_channels[-1] * 2

        stdv = 1.0 / math.sqrt(self.pool2d_avg_channels * 1.0)

        self.out = Linear(
            self.pool2d_avg_channels,
            class_dim,
            weight_attr=ParamAttr(
                initializer=Uniform(-stdv, stdv), name="fc6_weights"),
            bias_attr=ParamAttr(name="fc6_offset"))

    def forward(self, inputs):
         with paddle.static.amp.fp16_guard():
            if self.data_format == "NHWC":
                inputs = paddle.tensor.transpose(inputs, [0, 2, 3, 1])
                inputs.stop_gradient = True
            if self.layers < 152:
                y = self.conv(inputs)
            else:
                y = self.conv1_1(inputs)
                y = self.conv1_2(y)
                y = self.conv1_3(y)
            y = self.pool2d_max(y)
            for i, block in enumerate(self.block_list):
                y = block(y)
            y = self.pool2d_avg(y)
            y = paddle.reshape(y, shape=[-1, self.pool2d_avg_channels])
            y = self.out(y)
            return y


def SE_ResNeXt50_32x4d(**args):
    model = ResNeXt(layers=50, cardinality=32, **args)
    return model


def SE_ResNeXt101_32x4d(**args):
    model = ResNeXt(layers=101, cardinality=32, **args)
    return model


def SE_ResNeXt152_64x4d(**args):
    model = ResNeXt(layers=152, cardinality=64, **args)
    return model