|
|
|
|
|
|
|
|
""" |
|
|
@Author : Qingping Zheng |
|
|
@Contact : qingpingzheng2014@gmail.com |
|
|
@File : dml_csr.py |
|
|
@Time : 10/01/21 00:00 PM |
|
|
@Desc : |
|
|
@License : Licensed under the Apache License, Version 2.0 (the "License"); |
|
|
@Copyright : Copyright 2015 The Authors. All Rights Reserved. |
|
|
""" |
|
|
from __future__ import absolute_import |
|
|
from __future__ import division |
|
|
from __future__ import print_function |
|
|
|
|
|
|
|
|
import torch.nn as nn |
|
|
|
|
|
from torch.nn import functional as F |
|
|
from inplace_abn import InPlaceABNSync |
|
|
from .modules.ddgcn import DDualGCNHead |
|
|
from .modules.parsing import Parsing |
|
|
from .modules.edges import Edges |
|
|
from .modules.util import Bottleneck |
|
|
|
|
|
|
|
|
def conv3x3(in_planes, out_planes, stride=1): |
|
|
"3x3 convolution with padding" |
|
|
return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, |
|
|
padding=1, bias=False) |
|
|
|
|
|
|
|
|
class DML_CSR(nn.Module): |
|
|
def __init__(self, |
|
|
num_classes, |
|
|
abn=InPlaceABNSync, |
|
|
trained=True): |
|
|
super().__init__() |
|
|
self.inplanes = 128 |
|
|
self.is_trained = trained |
|
|
|
|
|
self.conv1 = conv3x3(3, 64, stride=2) |
|
|
self.bn1 = abn(64) |
|
|
self.relu1 = nn.ReLU(inplace=False) |
|
|
self.conv2 = conv3x3(64, 64) |
|
|
self.bn2 = abn(64) |
|
|
self.relu2 = nn.ReLU(inplace=False) |
|
|
self.conv3 = conv3x3(64, 128) |
|
|
self.bn3 = abn(128) |
|
|
self.relu3 = nn.ReLU(inplace=False) |
|
|
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1) |
|
|
self.layers = [3, 4, 23, 3] |
|
|
self.abn = abn |
|
|
strides = [1, 2, 1, 1] |
|
|
dilations = [1, 1, 1, 2] |
|
|
|
|
|
self.layer1 = self._make_layer(Bottleneck, 64, self.layers[0], stride=strides[0], dilation=dilations[0]) |
|
|
self.layer2 = self._make_layer(Bottleneck, 128, self.layers[1], stride=strides[1], dilation=dilations[1]) |
|
|
self.layer3 = self._make_layer(Bottleneck, 256, self.layers[2], stride=strides[2], dilation=dilations[2]) |
|
|
self.layer4 = self._make_layer(Bottleneck, 512, self.layers[3], stride=strides[3], dilation=dilations[3], multi_grid=(1,1,1)) |
|
|
|
|
|
self.context = DDualGCNHead(2048, 512, abn) |
|
|
self.layer6 = Parsing(512, 256, num_classes, abn) |
|
|
|
|
|
if self.is_trained: |
|
|
self.edge_layer = Edges(abn, out_fea=num_classes) |
|
|
|
|
|
def _make_layer(self, block, planes, blocks, stride=1, dilation=1, multi_grid=1): |
|
|
downsample = None |
|
|
if stride != 1 or self.inplanes != planes * block.expansion: |
|
|
downsample = nn.Sequential( |
|
|
nn.Conv2d(self.inplanes, planes * block.expansion, |
|
|
kernel_size=1, stride=stride, bias=False), |
|
|
self.abn(planes * block.expansion, affine=True)) |
|
|
|
|
|
layers = [] |
|
|
generate_multi_grid = lambda index, grids: grids[index%len(grids)] if isinstance(grids, tuple) else 1 |
|
|
layers.append(block(self.inplanes, planes, stride, abn=self.abn, dilation=dilation, downsample=downsample, multi_grid=generate_multi_grid(0, multi_grid))) |
|
|
self.inplanes = planes * block.expansion |
|
|
for i in range(1, blocks): |
|
|
layers.append(block(self.inplanes, planes, abn=self.abn, dilation=dilation, multi_grid=generate_multi_grid(i, multi_grid))) |
|
|
|
|
|
return nn.Sequential(*layers) |
|
|
|
|
|
def forward(self, x): |
|
|
input = x |
|
|
x = self.relu1(self.bn1(self.conv1(x))) |
|
|
x = self.relu2(self.bn2(self.conv2(x))) |
|
|
x1 = self.relu3(self.bn3(self.conv3(x))) |
|
|
x = self.maxpool(x1) |
|
|
x2 = self.layer1(x) |
|
|
x3 = self.layer2(x2) |
|
|
x4 = self.layer3(x3) |
|
|
x5 = self.layer4(x4) |
|
|
x = self.context(x5) |
|
|
seg, x = self.layer6(x, x2) |
|
|
|
|
|
if self.is_trained: |
|
|
binary_edge, semantic_edge, edge_fea = self.edge_layer(x2,x3,x4) |
|
|
return seg, binary_edge, semantic_edge |
|
|
|
|
|
return seg |
|
|
|
|
|
|
