IGEV/IGEV-Stereo/core/extractor.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from core.submodule import *
import timm




class ResidualBlock(nn.Module):
    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
        super(ResidualBlock, self).__init__()
  
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, padding=1, stride=stride)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, padding=1)
        self.relu = nn.ReLU(inplace=True)

        num_groups = planes // 8

        if norm_fn == 'group':
            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
            if not (stride == 1 and in_planes == planes):
                self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
        
        elif norm_fn == 'batch':
            self.norm1 = nn.BatchNorm2d(planes)
            self.norm2 = nn.BatchNorm2d(planes)
            if not (stride == 1 and in_planes == planes):
                self.norm3 = nn.BatchNorm2d(planes)
        
        elif norm_fn == 'instance':
            self.norm1 = nn.InstanceNorm2d(planes)
            self.norm2 = nn.InstanceNorm2d(planes)
            if not (stride == 1 and in_planes == planes):
                self.norm3 = nn.InstanceNorm2d(planes)

        elif norm_fn == 'none':
            self.norm1 = nn.Sequential()
            self.norm2 = nn.Sequential()
            if not (stride == 1 and in_planes == planes):
                self.norm3 = nn.Sequential()

        if stride == 1 and in_planes == planes:
            self.downsample = None
        
        else:    
            self.downsample = nn.Sequential(
                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm3)


    def forward(self, x):
        y = x
        y = self.conv1(y)
        y = self.norm1(y)
        y = self.relu(y)
        y = self.conv2(y)
        y = self.norm2(y)
        y = self.relu(y)

        if self.downsample is not None:
            x = self.downsample(x)

        return self.relu(x+y)



class BottleneckBlock(nn.Module):
    def __init__(self, in_planes, planes, norm_fn='group', stride=1):
        super(BottleneckBlock, self).__init__()
  
        self.conv1 = nn.Conv2d(in_planes, planes//4, kernel_size=1, padding=0)
        self.conv2 = nn.Conv2d(planes//4, planes//4, kernel_size=3, padding=1, stride=stride)
        self.conv3 = nn.Conv2d(planes//4, planes, kernel_size=1, padding=0)
        self.relu = nn.ReLU(inplace=True)

        num_groups = planes // 8

        if norm_fn == 'group':
            self.norm1 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
            self.norm2 = nn.GroupNorm(num_groups=num_groups, num_channels=planes//4)
            self.norm3 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
            if not stride == 1:
                self.norm4 = nn.GroupNorm(num_groups=num_groups, num_channels=planes)
        
        elif norm_fn == 'batch':
            self.norm1 = nn.BatchNorm2d(planes//4)
            self.norm2 = nn.BatchNorm2d(planes//4)
            self.norm3 = nn.BatchNorm2d(planes)
            if not stride == 1:
                self.norm4 = nn.BatchNorm2d(planes)
        
        elif norm_fn == 'instance':
            self.norm1 = nn.InstanceNorm2d(planes//4)
            self.norm2 = nn.InstanceNorm2d(planes//4)
            self.norm3 = nn.InstanceNorm2d(planes)
            if not stride == 1:
                self.norm4 = nn.InstanceNorm2d(planes)

        elif norm_fn == 'none':
            self.norm1 = nn.Sequential()
            self.norm2 = nn.Sequential()
            self.norm3 = nn.Sequential()
            if not stride == 1:
                self.norm4 = nn.Sequential()

        if stride == 1:
            self.downsample = None
        
        else:    
            self.downsample = nn.Sequential(
                nn.Conv2d(in_planes, planes, kernel_size=1, stride=stride), self.norm4)


    def forward(self, x):
        y = x
        y = self.relu(self.norm1(self.conv1(y)))
        y = self.relu(self.norm2(self.conv2(y)))
        y = self.relu(self.norm3(self.conv3(y)))

        if self.downsample is not None:
            x = self.downsample(x)

        return self.relu(x+y)

class BasicEncoder(nn.Module):
    def __init__(self, output_dim=128, norm_fn='batch', dropout=0.0, downsample=3):
        super(BasicEncoder, self).__init__()
        self.norm_fn = norm_fn
        self.downsample = downsample

        if self.norm_fn == 'group':
            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=64)
            
        elif self.norm_fn == 'batch':
            self.norm1 = nn.BatchNorm2d(64)

        elif self.norm_fn == 'instance':
            self.norm1 = nn.InstanceNorm2d(64)

        elif self.norm_fn == 'none':
            self.norm1 = nn.Sequential()

        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=1 + (downsample > 2), padding=3)
        self.relu1 = nn.ReLU(inplace=True)

        self.in_planes = 64
        self.layer1 = self._make_layer(64,  stride=1)
        self.layer2 = self._make_layer(96, stride=1 + (downsample > 1))
        self.layer3 = self._make_layer(128, stride=1 + (downsample > 0))

        # output convolution
        self.conv2 = nn.Conv2d(128, output_dim, kernel_size=1)

        self.dropout = None
        if dropout > 0:
            self.dropout = nn.Dropout2d(p=dropout)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
                if m.weight is not None:
                    nn.init.constant_(m.weight, 1)
                if m.bias is not None:
                    nn.init.constant_(m.bias, 0)

    def _make_layer(self, dim, stride=1):
        layer1 = ResidualBlock(self.in_planes, dim, self.norm_fn, stride=stride)
        layer2 = ResidualBlock(dim, dim, self.norm_fn, stride=1)
        layers = (layer1, layer2)
        
        self.in_planes = dim
        return nn.Sequential(*layers)


    def forward(self, x, dual_inp=False):

        # if input is list, combine batch dimension
        is_list = isinstance(x, tuple) or isinstance(x, list)
        if is_list:
            batch_dim = x[0].shape[0]
            x = torch.cat(x, dim=0)
        x = self.conv1(x)
        x = self.norm1(x)
        x = self.relu1(x)
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.conv2(x)

        if self.training and self.dropout is not None:
            x = self.dropout(x)

        if is_list:
            x = x.split(split_size=batch_dim, dim=0)

        return x

class MultiBasicEncoder(nn.Module):
    def __init__(self, output_dim=[128], norm_fn='batch', dropout=0.0, downsample=3):
        super(MultiBasicEncoder, self).__init__()
        self.norm_fn = norm_fn
        self.downsample = downsample

        # self.norm_111 = nn.BatchNorm2d(128, affine=False, track_running_stats=False)
        # self.norm_222 = nn.BatchNorm2d(128, affine=False, track_running_stats=False)

        if self.norm_fn == 'group':
            self.norm1 = nn.GroupNorm(num_groups=8, num_channels=64)

        elif self.norm_fn == 'batch':
            self.norm1 = nn.BatchNorm2d(64)

        elif self.norm_fn == 'instance':
            self.norm1 = nn.InstanceNorm2d(64)

        elif self.norm_fn == 'none':
            self.norm1 = nn.Sequential()

        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=1 + (downsample > 2), padding=3)
        self.relu1 = nn.ReLU(inplace=True)

        self.in_planes = 64
        self.layer1 = self._make_layer(64, stride=1)
        self.layer2 = self._make_layer(96, stride=1 + (downsample > 1))
        self.layer3 = self._make_layer(128, stride=1 + (downsample > 0))
        self.layer4 = self._make_layer(128, stride=2)
        self.layer5 = self._make_layer(128, stride=2)

        output_list = []
        
        for dim in output_dim:
            conv_out = nn.Sequential(
                ResidualBlock(128, 128, self.norm_fn, stride=1),
                nn.Conv2d(128, dim[2], 3, padding=1))
            output_list.append(conv_out)

        self.outputs04 = nn.ModuleList(output_list)

        output_list = []
        for dim in output_dim:
            conv_out = nn.Sequential(
                ResidualBlock(128, 128, self.norm_fn, stride=1),
                nn.Conv2d(128, dim[1], 3, padding=1))
            output_list.append(conv_out)

        self.outputs08 = nn.ModuleList(output_list)

        output_list = []
        for dim in output_dim:
            conv_out = nn.Conv2d(128, dim[0], 3, padding=1)
            output_list.append(conv_out)

        self.outputs16 = nn.ModuleList(output_list)

        if dropout > 0:
            self.dropout = nn.Dropout2d(p=dropout)
        else:
            self.dropout = None

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d, nn.GroupNorm)):
                if m.weight is not None:
                    nn.init.constant_(m.weight, 1)
                if m.bias is not None:
                    nn.init.constant_(m.bias, 0)

    def _make_layer(self, dim, stride=1):
        layer1 = ResidualBlock(self.in_planes, dim, self.norm_fn, stride=stride)
        layer2 = ResidualBlock(dim, dim, self.norm_fn, stride=1)
        layers = (layer1, layer2)

        self.in_planes = dim
        return nn.Sequential(*layers)

    def forward(self, x, dual_inp=False, num_layers=3):

        x = self.conv1(x)
        x = self.norm1(x)
        x = self.relu1(x)
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        if dual_inp:
            v = x
            x = x[:(x.shape[0]//2)]

        outputs04 = [f(x) for f in self.outputs04]
        if num_layers == 1:
            return (outputs04, v) if dual_inp else (outputs04,)

        y = self.layer4(x)
        outputs08 = [f(y) for f in self.outputs08]

        if num_layers == 2:
            return (outputs04, outputs08, v) if dual_inp else (outputs04, outputs08)

        z = self.layer5(y)
        outputs16 = [f(z) for f in self.outputs16]

        return (outputs04, outputs08, outputs16, v) if dual_inp else (outputs04, outputs08, outputs16)


class SubModule(nn.Module):
    def __init__(self):
        super(SubModule, self).__init__()

    def weight_init(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
                m.weight.data.normal_(0, math.sqrt(2. / n))
            elif isinstance(m, nn.Conv3d):
                n = m.kernel_size[0] * m.kernel_size[1] * m.kernel_size[2] * m.out_channels
                m.weight.data.normal_(0, math.sqrt(2. / n))
            elif isinstance(m, nn.BatchNorm2d):
                m.weight.data.fill_(1)
                m.bias.data.zero_()
            elif isinstance(m, nn.BatchNorm3d):
                m.weight.data.fill_(1)
                m.bias.data.zero_()


class Feature(SubModule):
    def __init__(self, freeze):
        super(Feature, self).__init__()
        pretrained =  True
        self.model = timm.create_model('mobilenetv2_100', pretrained=pretrained, features_only=True)
        if freeze:
            for p in self.model.parameters():
                p.requires_grad = False
        layers = [1,2,3,5,6]
        chans = [16, 24, 32, 96, 160]
        self.conv_stem = self.model.conv_stem
        self.bn1 = self.model.bn1
        self.act1 = self.model.act1

        self.block0 = torch.nn.Sequential(*self.model.blocks[0:layers[0]])
        self.block1 = torch.nn.Sequential(*self.model.blocks[layers[0]:layers[1]])
        self.block2 = torch.nn.Sequential(*self.model.blocks[layers[1]:layers[2]])
        self.block3 = torch.nn.Sequential(*self.model.blocks[layers[2]:layers[3]])
        self.block4 = torch.nn.Sequential(*self.model.blocks[layers[3]:layers[4]])

        self.deconv32_16 = Conv2x_IN(chans[4], chans[3], deconv=True, concat=True)
        self.deconv16_8 = Conv2x_IN(chans[3]*2, chans[2], deconv=True, concat=True)
        self.deconv8_4 = Conv2x_IN(chans[2]*2, chans[1], deconv=True, concat=True)
        self.conv4 = BasicConv_IN(chans[1]*2, chans[1]*2, kernel_size=3, stride=1, padding=1)

    def forward(self, x):
        x = self.act1(self.bn1(self.conv_stem(x)))
        x2 = self.block0(x)
        x4 = self.block1(x2)
        x8 = self.block2(x4)
        x16 = self.block3(x8)
        x32 = self.block4(x16)

        x16 = self.deconv32_16(x32, x16)
        x8 = self.deconv16_8(x16, x8)
        x4 = self.deconv8_4(x8, x4)
        x4 = self.conv4(x4)
        return [x4, x8, x16, x32]