paddleseg/models/mobileseg.py

# copyright (c) 2022 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.

import paddle
import paddle.nn as nn
import paddle.nn.functional as F

from paddleseg import utils
from paddleseg.models import layers
from paddleseg.cvlibs import manager


@manager.MODELS.add_component
class MobileSeg(nn.Layer):
    """
    The semantic segmentation models for mobile devices.

    Args:
        num_classes (int): The number of target classes.
        backbone(nn.Layer): Backbone network, such as stdc1net and resnet18. The backbone must
            has feat_channels, of which the length is 5.
        backbone_indices (List(int), optional): The values indicate the indices of output of backbone.
            Default: [2, 3, 4].
        cm_bin_sizes (List(int), optional): The bin size of context module. Default: [1,2,4].
        cm_out_ch (int, optional): The output channel of the last context module. Default: 128.
        arm_type (str, optional): The type of attention refinement module. Default: ARM_Add_SpAttenAdd3.
        arm_out_chs (List(int), optional): The out channels of each arm module. Default: [64, 96, 128].
        seg_head_inter_chs (List(int), optional): The intermediate channels of segmentation head.
            Default: [64, 64, 64].
        resize_mode (str, optional): The resize mode for the upsampling operation in decoder.
            Default: bilinear.
        use_last_fuse (bool, optional): Whether use fusion in the last. Default: False.
        pretrained (str, optional): The path or url of pretrained model. Default: None.
    """

    def __init__(self,
                 num_classes,
                 backbone,
                 backbone_indices=[1, 2, 3],
                 cm_bin_sizes=[1, 2],
                 cm_out_ch=64,
                 arm_type='UAFMMobile',
                 arm_out_chs=[32, 48, 64],
                 seg_head_inter_chs=[32, 32, 32],
                 resize_mode='bilinear',
                 use_last_fuse=False,
                 pretrained=None):
        super().__init__()

        # backbone
        assert hasattr(backbone, 'feat_channels'), \
            "The backbone should has feat_channels."
        assert len(backbone.feat_channels) >= len(backbone_indices), \
            f"The length of input backbone_indices ({len(backbone_indices)}) should not be" \
            f"greater than the length of feat_channels ({len(backbone.feat_channels)})."
        assert len(backbone.feat_channels) > max(backbone_indices), \
            f"The max value ({max(backbone_indices)}) of backbone_indices should be " \
            f"less than the length of feat_channels ({len(backbone.feat_channels)})."
        self.backbone = backbone

        assert len(backbone_indices) >= 1, "The lenght of backbone_indices " \
            "should not be lesser than 1"
        self.backbone_indices = backbone_indices  # [..., x16_id, x32_id]
        backbone_out_chs = [backbone.feat_channels[i] for i in backbone_indices]

        # head
        if len(arm_out_chs) == 1:
            arm_out_chs = arm_out_chs * len(backbone_indices)
        assert len(arm_out_chs) == len(backbone_indices), "The length of " \
            "arm_out_chs and backbone_indices should be equal"

        self.ppseg_head = MobileSegHead(backbone_out_chs, arm_out_chs,
                                        cm_bin_sizes, cm_out_ch, arm_type,
                                        resize_mode, use_last_fuse)

        if len(seg_head_inter_chs) == 1:
            seg_head_inter_chs = seg_head_inter_chs * len(backbone_indices)
        assert len(seg_head_inter_chs) == len(backbone_indices), "The length of " \
            "seg_head_inter_chs and backbone_indices should be equal"
        self.seg_heads = nn.LayerList()  # [..., head_16, head32]
        for in_ch, mid_ch in zip(arm_out_chs, seg_head_inter_chs):
            self.seg_heads.append(SegHead(in_ch, mid_ch, num_classes))

        # pretrained
        self.pretrained = pretrained
        self.init_weight()

    def forward(self, x):
        x_hw = paddle.shape(x)[2:]

        feats_backbone = self.backbone(x)  # [x4, x8, x16, x32]
        assert len(feats_backbone) >= len(self.backbone_indices), \
            f"The nums of backbone feats ({len(feats_backbone)}) should be greater or " \
            f"equal than the nums of backbone_indices ({len(self.backbone_indices)})"

        feats_selected = [feats_backbone[i] for i in self.backbone_indices]
        feats_head = self.ppseg_head(feats_selected)  # [..., x8, x16, x32]

        if self.training:
            logit_list = []
            for x, seg_head in zip(feats_head, self.seg_heads):
                x = seg_head(x)
                logit_list.append(x)
            logit_list = [
                F.interpolate(
                    x, x_hw, mode='bilinear', align_corners=False)
                for x in logit_list
            ]
        else:
            x = self.seg_heads[0](feats_head[0])
            x = F.interpolate(x, x_hw, mode='bilinear', align_corners=False)
            logit_list = [x]

        return logit_list

    def init_weight(self):
        if self.pretrained is not None:
            utils.load_entire_model(self, self.pretrained)


class MobileSegHead(nn.Layer):
    """
    The head of MobileSeg.

    Args:
        backbone_out_chs (List(Tensor)): The channels of output tensors in the backbone.
        arm_out_chs (List(int)): The out channels of each arm module.
        cm_bin_sizes (List(int)): The bin size of context module.
        cm_out_ch (int): The output channel of the last context module.
        arm_type (str): The type of attention refinement module.
        resize_mode (str): The resize mode for the upsampling operation in decoder.
    """

    def __init__(self, backbone_out_chs, arm_out_chs, cm_bin_sizes, cm_out_ch,
                 arm_type, resize_mode, use_last_fuse):
        super().__init__()

        self.cm = MobileContextModule(backbone_out_chs[-1], cm_out_ch,
                                      cm_out_ch, cm_bin_sizes)

        assert hasattr(layers,arm_type), \
            "Not support arm_type ({})".format(arm_type)
        arm_class = eval("layers." + arm_type)

        self.arm_list = nn.LayerList()  # [..., arm8, arm16, arm32]
        for i in range(len(backbone_out_chs)):
            low_chs = backbone_out_chs[i]
            high_ch = cm_out_ch if i == len(
                backbone_out_chs) - 1 else arm_out_chs[i + 1]
            out_ch = arm_out_chs[i]
            arm = arm_class(
                low_chs, high_ch, out_ch, ksize=3, resize_mode=resize_mode)
            self.arm_list.append(arm)

        self.use_last_fuse = use_last_fuse
        if self.use_last_fuse:
            self.fuse_convs = nn.LayerList()
            for i in range(1, len(arm_out_chs)):
                conv = layers.SeparableConvBNReLU(
                    arm_out_chs[i],
                    arm_out_chs[0],
                    kernel_size=3,
                    bias_attr=False)
                self.fuse_convs.append(conv)
            self.last_conv = layers.SeparableConvBNReLU(
                len(arm_out_chs) * arm_out_chs[0],
                arm_out_chs[0],
                kernel_size=3,
                bias_attr=False)

    def forward(self, in_feat_list):
        """
        Args:
            in_feat_list (List(Tensor)): Such as [x2, x4, x8, x16, x32].
                x2, x4 and x8 are optional.
        Returns:
            out_feat_list (List(Tensor)): Such as [x2, x4, x8, x16, x32].
                x2, x4 and x8 are optional.
                The length of in_feat_list and out_feat_list are the same.
        """

        high_feat = self.cm(in_feat_list[-1])
        out_feat_list = []

        for i in reversed(range(len(in_feat_list))):
            low_feat = in_feat_list[i]
            arm = self.arm_list[i]
            high_feat = arm(low_feat, high_feat)
            out_feat_list.insert(0, high_feat)

        if self.use_last_fuse:
            x_list = [out_feat_list[0]]
            size = paddle.shape(out_feat_list[0])[2:]
            for i, (x, conv
                    ) in enumerate(zip(out_feat_list[1:], self.fuse_convs)):
                x = conv(x)
                x = F.interpolate(
                    x, size=size, mode='bilinear', align_corners=False)
                x_list.append(x)
            x = paddle.concat(x_list, axis=1)
            x = self.last_conv(x)
            out_feat_list[0] = x

        return out_feat_list


class MobileContextModule(nn.Layer):
    """
    Context Module for Mobile Model.

    Args:
        in_channels (int): The number of input channels to pyramid pooling module.
        inter_channels (int): The number of inter channels to pyramid pooling module.
        out_channels (int): The number of output channels after pyramid pooling module.
        bin_sizes (tuple, optional): The out size of pooled feature maps. Default: (1, 3).
        align_corners (bool): An argument of F.interpolate. It should be set to False
            when the output size of feature is even, e.g. 1024x512, otherwise it is True, e.g. 769x769.
    """

    def __init__(self,
                 in_channels,
                 inter_channels,
                 out_channels,
                 bin_sizes,
                 align_corners=False):
        super().__init__()

        self.stages = nn.LayerList([
            self._make_stage(in_channels, inter_channels, size)
            for size in bin_sizes
        ])

        self.conv_out = layers.SeparableConvBNReLU(
            in_channels=inter_channels,
            out_channels=out_channels,
            kernel_size=3,
            bias_attr=False)

        self.align_corners = align_corners

    def _make_stage(self, in_channels, out_channels, size):
        prior = nn.AdaptiveAvgPool2D(output_size=size)
        conv = layers.ConvBNReLU(
            in_channels=in_channels, out_channels=out_channels, kernel_size=1)
        return nn.Sequential(prior, conv)

    def forward(self, input):
        out = None
        input_shape = paddle.shape(input)[2:]

        for stage in self.stages:
            x = stage(input)
            x = F.interpolate(
                x,
                input_shape,
                mode='bilinear',
                align_corners=self.align_corners)
            if out is None:
                out = x
            else:
                out += x

        out = self.conv_out(out)
        return out


class SegHead(nn.Layer):
    def __init__(self, in_chan, mid_chan, n_classes):
        super().__init__()
        self.conv = layers.SeparableConvBNReLU(
            in_chan, mid_chan, kernel_size=3, bias_attr=False)
        self.conv_out = nn.Conv2D(
            mid_chan, n_classes, kernel_size=1, bias_attr=False)

    def forward(self, x):
        x = self.conv(x)
        x = self.conv_out(x)
        return x