mmdet/models/backbones/mobilenet_v2.py

# Copyright (c) OpenMMLab. All rights reserved.
import warnings

import torch.nn as nn
from mmcv.cnn import ConvModule
from mmengine.model import BaseModule
from torch.nn.modules.batchnorm import _BatchNorm

from mmdet.registry import MODELS
from ..layers import InvertedResidual
from ..utils import make_divisible


@MODELS.register_module()
class MobileNetV2(BaseModule):
    """MobileNetV2 backbone.

    Args:
        widen_factor (float): Width multiplier, multiply number of
            channels in each layer by this amount. Default: 1.0.
        out_indices (Sequence[int], optional): Output from which stages.
            Default: (1, 2, 4, 7).
        frozen_stages (int): Stages to be frozen (all param fixed).
            Default: -1, which means not freezing any parameters.
        conv_cfg (dict, optional): Config dict for convolution layer.
            Default: None, which means using conv2d.
        norm_cfg (dict): Config dict for normalization layer.
            Default: dict(type='BN').
        act_cfg (dict): Config dict for activation layer.
            Default: dict(type='ReLU6').
        norm_eval (bool): Whether to set norm layers to eval mode, namely,
            freeze running stats (mean and var). Note: Effect on Batch Norm
            and its variants only. Default: False.
        with_cp (bool): Use checkpoint or not. Using checkpoint will save some
            memory while slowing down the training speed. Default: False.
        pretrained (str, optional): model pretrained path. Default: None
        init_cfg (dict or list[dict], optional): Initialization config dict.
            Default: None
    """

    # Parameters to build layers. 4 parameters are needed to construct a
    # layer, from left to right: expand_ratio, channel, num_blocks, stride.
    arch_settings = [[1, 16, 1, 1], [6, 24, 2, 2], [6, 32, 3, 2],
                     [6, 64, 4, 2], [6, 96, 3, 1], [6, 160, 3, 2],
                     [6, 320, 1, 1]]

    def __init__(self,
                 widen_factor=1.,
                 out_indices=(1, 2, 4, 7),
                 frozen_stages=-1,
                 conv_cfg=None,
                 norm_cfg=dict(type='BN'),
                 act_cfg=dict(type='ReLU6'),
                 norm_eval=False,
                 with_cp=False,
                 pretrained=None,
                 init_cfg=None):
        super(MobileNetV2, self).__init__(init_cfg)

        self.pretrained = pretrained
        assert not (init_cfg and pretrained), \
            'init_cfg and pretrained cannot be specified at the same time'
        if isinstance(pretrained, str):
            warnings.warn('DeprecationWarning: pretrained is deprecated, '
                          'please use "init_cfg" instead')
            self.init_cfg = dict(type='Pretrained', checkpoint=pretrained)
        elif pretrained is None:
            if init_cfg is None:
                self.init_cfg = [
                    dict(type='Kaiming', layer='Conv2d'),
                    dict(
                        type='Constant',
                        val=1,
                        layer=['_BatchNorm', 'GroupNorm'])
                ]
        else:
            raise TypeError('pretrained must be a str or None')

        self.widen_factor = widen_factor
        self.out_indices = out_indices
        if not set(out_indices).issubset(set(range(0, 8))):
            raise ValueError('out_indices must be a subset of range'
                             f'(0, 8). But received {out_indices}')

        if frozen_stages not in range(-1, 8):
            raise ValueError('frozen_stages must be in range(-1, 8). '
                             f'But received {frozen_stages}')
        self.out_indices = out_indices
        self.frozen_stages = frozen_stages
        self.conv_cfg = conv_cfg
        self.norm_cfg = norm_cfg
        self.act_cfg = act_cfg
        self.norm_eval = norm_eval
        self.with_cp = with_cp

        self.in_channels = make_divisible(32 * widen_factor, 8)

        self.conv1 = ConvModule(
            in_channels=3,
            out_channels=self.in_channels,
            kernel_size=3,
            stride=2,
            padding=1,
            conv_cfg=self.conv_cfg,
            norm_cfg=self.norm_cfg,
            act_cfg=self.act_cfg)

        self.layers = []

        for i, layer_cfg in enumerate(self.arch_settings):
            expand_ratio, channel, num_blocks, stride = layer_cfg
            out_channels = make_divisible(channel * widen_factor, 8)
            inverted_res_layer = self.make_layer(
                out_channels=out_channels,
                num_blocks=num_blocks,
                stride=stride,
                expand_ratio=expand_ratio)
            layer_name = f'layer{i + 1}'
            self.add_module(layer_name, inverted_res_layer)
            self.layers.append(layer_name)

        if widen_factor > 1.0:
            self.out_channel = int(1280 * widen_factor)
        else:
            self.out_channel = 1280

        layer = ConvModule(
            in_channels=self.in_channels,
            out_channels=self.out_channel,
            kernel_size=1,
            stride=1,
            padding=0,
            conv_cfg=self.conv_cfg,
            norm_cfg=self.norm_cfg,
            act_cfg=self.act_cfg)
        self.add_module('conv2', layer)
        self.layers.append('conv2')

    def make_layer(self, out_channels, num_blocks, stride, expand_ratio):
        """Stack InvertedResidual blocks to build a layer for MobileNetV2.

        Args:
            out_channels (int): out_channels of block.
            num_blocks (int): number of blocks.
            stride (int): stride of the first block. Default: 1
            expand_ratio (int): Expand the number of channels of the
                hidden layer in InvertedResidual by this ratio. Default: 6.
        """
        layers = []
        for i in range(num_blocks):
            if i >= 1:
                stride = 1
            layers.append(
                InvertedResidual(
                    self.in_channels,
                    out_channels,
                    mid_channels=int(round(self.in_channels * expand_ratio)),
                    stride=stride,
                    with_expand_conv=expand_ratio != 1,
                    conv_cfg=self.conv_cfg,
                    norm_cfg=self.norm_cfg,
                    act_cfg=self.act_cfg,
                    with_cp=self.with_cp))
            self.in_channels = out_channels

        return nn.Sequential(*layers)

    def _freeze_stages(self):
        if self.frozen_stages >= 0:
            for param in self.conv1.parameters():
                param.requires_grad = False
        for i in range(1, self.frozen_stages + 1):
            layer = getattr(self, f'layer{i}')
            layer.eval()
            for param in layer.parameters():
                param.requires_grad = False

    def forward(self, x):
        """Forward function."""
        x = self.conv1(x)
        outs = []
        for i, layer_name in enumerate(self.layers):
            layer = getattr(self, layer_name)
            x = layer(x)
            if i in self.out_indices:
                outs.append(x)
        return tuple(outs)

    def train(self, mode=True):
        """Convert the model into training mode while keep normalization layer
        frozen."""
        super(MobileNetV2, self).train(mode)
        self._freeze_stages()
        if mode and self.norm_eval:
            for m in self.modules():
                # trick: eval have effect on BatchNorm only
                if isinstance(m, _BatchNorm):
                    m.eval()