ECBBBBBBBBB

最新推荐文章于 2024-10-15 20:30:18 发布

superbestao

最新推荐文章于 2024-10-15 20:30:18 发布

阅读量113

点赞数

文章标签：深度学习 python pytorch

本文链接：https://blog.csdn.net/weixin_41595822/article/details/126700254

版权

该博客详细介绍了PFPN（Progressive Feature Fusion Network）的实现，包括其核心模块如ResNet101、FeaturePolishingModule、ECA模块等。PFPN在网络中用于特征融合和精细化，通过不同尺度的特征融合提升模型性能。同时，博客还展示了代码实现，涉及卷积、批量归一化、全局平均池化等操作。

摘要由CSDN通过智能技术生成

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# Copyright (C) 2019 Alibaba Group Holding Limited.
# All rights reserved.
# 
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
# 1. Redistributions of source code must retain the above copyright
#    notice, this list of conditions and the following disclaimer.
# 2. Redistributions in binary form must reproduce the above copyright
#    notice, this list of conditions and the following disclaimer in the
#    documentation and/or other materials provided with the distribution.
# 
# THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
# ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
# OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
# HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
# OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
# SUCH DAMAGE.
##############################################################################

import torch
from torch import nn
from torch.nn import init
import torch.nn.functional as F
from torch.autograd import Variable
import torchvision
import numpy as np
import cv2
import sys
import os
import math

from networks.layers import layer_weights_init, BatchNormFixed
from networks.backbone import ResNet, ResNet101
from networks.config import *

__all__ = ['PFPN',
        ]

Normal_Type = 'batch'
Momentum = 0.1
def _normal_layer(in_ch):
    if Normal_Type == 'fixed_bn':
        return BatchNormFixed(in_ch)
    elif Normal_Type == 'group':
        return nn.GroupNorm(in_ch, in_ch)
    else:
        return nn.BatchNorm2d(in_ch, momentum=Momentum)


class PFPN(nn.Module):
    def __init__(self, load_pretrained=True):
        super(PFPN, self).__init__()
        if load_pretrained:
            self.pretrained = model_addr['res101']
        else:
            self.pretrained = None
        self.base = PFPNBase(fix_bn=True,
                           strides=[1, 2, 2, 2],
                           dilations=[1, 1, 1, 1],
                           pretrained=self.pretrained)
        branch_planes = [256, 256, 256, 256, 256]
        self.transition = Transition(
            inplanes=[64, 256, 512, 1024, 2048],
            outplanes=branch_planes)
        self.fpm1 = FeaturePolishingModule(
            inplanes=branch_planes,
            outplanes=branch_planes,
            size_times=[1, 2, 4, 8, 16])
        self.fpm2 = FeaturePolishingModule(
            inplanes=branch_planes,
            outplanes=branch_planes,
            size_times=[1, 2, 4, 8, 16])
        self.fpm3 = FeaturePolishingModule(
            inplanes=branch_planes,
            outplanes=branch_planes,
            size_times=[1, 2, 4, 8, 16])