简介：
1.该教程提供大量的首发改进的方式，降低上手难度，多种结构改进，助力寻找创新点！
2.本篇文章对Pointnet++进行激活函数的改进，助力解决RELU激活函数缺陷。
3.专栏持续更新，紧随最新的研究内容。
文章目录
- 步骤一
- 步骤二
- 步骤三
代码地址
步骤一
新建activate.py文件，我存放在新建的block目录下，加入以下代码：
# Activation functionsimport torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np# SiLU https://arxiv.org/pdf/1606.08415.pdf ----------------------------------------------------------------------------
class SiLU(nn.Module): # export-friendly version of nn.SiLU()@staticmethoddef forward(x):return x * torch.sigmoid(x)class Hardswish(nn.Module): # export-friendly version of nn.Hardswish()@staticmethoddef forward(x):# return x * F.hardsigmoid(x) # for torchscript and CoreMLreturn x * F.hardtanh(x + 3, 0., 6.) / 6. # for torchscript, CoreML and ONNXclass MemoryEfficientSwish(nn.Module):class F(torch.autograd.Function):@staticmethoddef forward(ctx, x):ctx.save_for_backward(x)return x * torch.sigmoid(x)@staticmethoddef backward(ctx, grad_output):x = ctx.saved_tensors[0]sx = torch.sigmoid(x)return grad_output * (sx * (1 + x * (1 - sx)))def forward(self, x):return self.F.apply(x)# Mish https://github.com/digantamisra98/Mish --------------------------------------------------------------------------
class Mish(nn.Module):@staticmethoddef forward(x):return x * F.softplus(x).tanh()class MemoryEfficientMish(nn.Module):class F(torch.autograd.Function):@staticmethoddef forward(ctx, x):ctx.save_for_backward(x)return x.mul(torch.tanh(F.softplus(x))) # x * tanh(ln(1 + aconcxunlian(x)))@staticmethoddef backward(ctx, grad_output):x = ctx.saved_tensors[0]sx = torch.sigmoid(x)fx = F.softplus(x).tanh()return grad_output * (fx + x * sx * (1 - fx * fx))def forward(self, x):return self.F.apply(x)# FReLU https://arxiv.org/abs/2007.11824 -------------------------------------------------------------------------------
class FReLU(nn.Module):def __init__(self, c1, k=3): # ch_in, kernelsuper().__init__()self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False)self.bn = nn.BatchNorm2d(c1)def forward(self, x):return torch.max(x, self.bn(self.conv(x)))class GELU(nn.Module):def __init__(self):super(GELU, self).__init__()def forward(self, x):return 0.5 * x * (1 + torch.tanh(np.sqrt(2 / np.pi) * (x + 0.044715 * torch.pow(x, 3))))#
class MetaAconC(nn.Module):r""" ACON activation (activate or not).MetaAconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is generated by a small networkaccording to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>."""def __init__(self, c1, k=1, s=1, r=16): # ch_in, kernel, stride, rsuper().__init__()c2 = max(r, c1 // r)self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1))self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1))self.fc1 = nn.Conv2d(c1, c2, k, s, bias=True)self.fc2 = nn.Conv2d(c2, c1, k, s, bias=True)# self.bn1 = nn.BatchNorm2d(c2)# self.bn2 = nn.BatchNorm2d(c1)def forward(self, x):y = x.mean(dim=2, keepdims=True).mean(dim=3, keepdims=True)# batch-size 1 bug/instabilities https://github.com/ultralytics/yolov5/issues/2891# beta = torch.sigmoid(self.bn2(self.fc2(self.bn1(self.fc1(y))))) # bug/unstablebeta = torch.sigmoid(self.fc2(self.fc1(y))) # bug patch BN layers removeddpx = (self.p1 - self.p2) * xreturn dpx * torch.sigmoid(beta * dpx) + self.p2 * x
###
class AconC(nn.Module):"""ACON https://arxiv.org/pdf/2009.04759.pdfACON activation (activate or not).AconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is a learnable parameteraccording to "Activate or Not: Learning Customized Activation" <https://arxiv.org/pdf/2009.04759.pdf>."""def __init__(self, c1):super().__init__()self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1))self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1))self.beta = nn.Parameter(torch.ones(1, c1, 1, 1))def forward(self, x):dpx = (self.p1 - self.p2) * xreturn dpx * torch.sigmoid(self.beta * dpx) + self.p2 * x步骤二
在models/pointnet2_utils.py中加入以下代码，该代码将PointNetSetAbstraction中的mlp三层感知机重新封装成一个class Conv模块，便于直接在Conv模块中修改激活函数，修改后的代码和源码结构是一致的。修改不同的激活函数直接在Conv类中修改即可。
PointNetSetAbstraction结构图如下，PointNetSetAbstractionMSG比PointNetSetAbstraction多一个不同尺度的三层mlp,其他结构是一样的。
class Conv(nn.Module):# Standard convolutiondef __init__(self, c1, c2, k=1): # ch_in, ch_out, kernel, stride, padding, groupssuper(Conv, self).__init__()self.conv = nn.Conv2d(c1, c2, k)self.bn = nn.BatchNorm2d(c2)#self.act = nn.SiLU()#self.act = nn.LeakyReLU(0.1)self.act = nn.ReLU()#self.act = MetaAconC(c2)#self.act = AconC(c2)#self.act = Mish()#self.act = Hardswish()#self.act = FReLU(c2)def forward(self, x):return self.act(self.bn(self.conv(x)))def fuseforward(self, x):return self.act(self.conv(x))class PointNetSetAbstractionAttention(nn.Module):def __init__(self, npoint, radius, nsample, in_channel, mlp, group_all):super(PointNetSetAbstractionAttention, self).__init__()self.npoint = npointself.radius = radiusself.nsample = nsample#self.mlp_convs = nn.ModuleList()self.mlp_conv1 = Conv(in_channel,mlp[0],1)self.mlp_attention = CBAM(mlp[0])self.mlp_conv2 = Conv(mlp[0],mlp[1],1)self.mlp_conv3 = Conv(mlp[1],mlp[2],1)self.group_all = group_alldef forward(self, xyz, points):"""Input:xyz: input points position data, [B, C, N]points: input points data, [B, D, N]Return:new_xyz: sampled points position data, [B, C, S]new_points_concat: sample points feature data, [B, D', S]"""xyz = xyz.permute(0, 2, 1)if points is not None:points = points.permute(0, 2, 1)if self.group_all:new_xyz, new_points = sample_and_group_all(xyz, points)else:new_xyz, new_points = sample_and_group(self.npoint, self.radius, self.nsample, xyz, points)# new_xyz: sampled points position data, [B, npoint, C]# new_points: sampled points data, [B, npoint, nsample, C+D]new_points = new_points.permute(0, 3, 2, 1) # [B, C+D, nsample,npoint]new_points=self.mlp_conv1(new_points)new_points = self.mlp_attention(new_points)new_points = self.mlp_conv2(new_points)new_points = self.mlp_conv3(new_points)new_points = torch.max(new_points, 2)[0]new_xyz = new_xyz.permute(0, 2, 1)return new_xyz, new_pointsclass PointNetSetAbstractionMsgAttention(nn.Module):def __init__(self, npoint, radius_list, nsample_list, in_channel, mlp_list):super(PointNetSetAbstractionMsgAttention, self).__init__()self.npoint = npointself.radius_list = radius_listself.nsample_list = nsample_listself.mlp_conv00 = Conv(in_channel+3,mlp_list[0][0],1)self.mlp_conv01 = Conv(mlp_list[0][0],mlp_list[0][1],1)self.mlp_conv02 = Conv(mlp_list[0][1],mlp_list[0][2],1)self.mlp_conv10 = Conv(in_channel+3,mlp_list[1][0],1)self.mlp_conv11 = Conv(mlp_list[1][0],mlp_list[1][1],1)self.mlp_conv12 = Conv(mlp_list[1][1],mlp_list[1][2],1)# self.conv_blocks = nn.ModuleList()# self.bn_blocks = nn.ModuleList()# for i in range(len(mlp_list)):# convs = nn.ModuleList()# bns = nn.ModuleList()# last_channel = in_channel + 3# for out_channel in mlp_list[i]:# convs.append(nn.Conv2d(last_channel, out_channel, 1))# bns.append(nn.BatchNorm2d(out_channel))# last_channel = out_channel# self.conv_blocks.append(convs)# self.bn_blocks.append(bns)def forward(self, xyz, points):"""Input:xyz: input points position data, [B, C, N]points: input points data, [B, D, N]Return:new_xyz: sampled points position data, [B, C, S]new_points_concat: sample points feature data, [B, D', S]"""xyz = xyz.permute(0, 2, 1)if points is not None:points = points.permute(0, 2, 1)B, N, C = xyz.shapeS = self.npointnew_xyz = index_points(xyz, farthest_point_sample(xyz, S))new_points_list = []for i, radius in enumerate(self.radius_list):K = self.nsample_list[i]group_idx = query_ball_point(radius, K, xyz, new_xyz)grouped_xyz = index_points(xyz, group_idx)grouped_xyz -= new_xyz.view(B, S, 1, C)if points is not None:grouped_points = index_points(points, group_idx)grouped_points = torch.cat([grouped_points, grouped_xyz], dim=-1)else:grouped_points = grouped_xyzgrouped_points = grouped_points.permute(0, 3, 2, 1) # [B, D, K, S]if i==0:grouped_points =self.mlp_conv00(grouped_points)grouped_points = self.mlp_conv01(grouped_points)grouped_points = self.mlp_conv02(grouped_points)else:grouped_points = self.mlp_conv10(grouped_points)grouped_points = self.mlp_conv11(grouped_points)grouped_points = self.mlp_conv12(grouped_points)# for j in range(len(self.conv_blocks[i])):# conv = self.conv_blocks[i][j]# bn = self.bn_blocks[i][j]# grouped_points = F.relu(bn(conv(grouped_points)))new_points = torch.max(grouped_points, 2)[0] # [B, D', S]new_points_list.append(new_points)new_xyz = new_xyz.permute(0, 2, 1)new_points_concat = torch.cat(new_points_list, dim=1)return new_xyz, new_points_concat步骤三
在不同的模型中修改调用即可，如在models/pointnet2_sem_seg.py文件中修改，训练即可
import torch.nn as nn
import torch.nn.functional as F
# from models.pointnet2_utils import PointNetSetAbstraction, PointNetFeaturePropagation, PointNetSetAbstractionKPconv, \
# PointNetSetAbstractionAttention
from models.pointnet2_utils import *class get_model(nn.Module):def __init__(self, num_classes):super(get_model, self).__init__()self.sa1 = PointNetSetAbstractionAttention(1024, 0.1, 32, 9 + 3, [32, 32, 64], False)self.sa2 = PointNetSetAbstraction(256, 0.2, 32, 64 + 3, [64, 64, 128], False)self.sa3 = PointNetSetAbstraction(64, 0.4, 32, 128 + 3, [128, 128, 256], False)self.sa4 = PointNetSetAbstraction(16, 0.8, 32, 256 + 3, [256, 256, 512], False)self.fp4 = PointNetFeaturePropagation(768, [256, 256])self.fp3 = PointNetFeaturePropagation(384, [256, 256])self.fp2 = PointNetFeaturePropagation(320, [256, 128])self.fp1 = PointNetFeaturePropagation(128, [128, 128, 128])self.conv1 = nn.Conv1d(128, 128, 1)self.bn1 = nn.BatchNorm1d(128)self.drop1 = nn.Dropout(0.5)self.conv2 = nn.Conv1d(128, num_classes, 1)def forward(self, xyz):l0_points = xyzl0_xyz = xyz[:,:3,:]l1_xyz, l1_points = self.sa1(l0_xyz, l0_points)l2_xyz, l2_points = self.sa2(l1_xyz, l1_points)l3_xyz, l3_points = self.sa3(l2_xyz, l2_points)l4_xyz, l4_points = self.sa4(l3_xyz, l3_points)l3_points = self.fp4(l3_xyz, l4_xyz, l3_points, l4_points)l2_points = self.fp3(l2_xyz, l3_xyz, l2_points, l3_points)l1_points = self.fp2(l1_xyz, l2_xyz, l1_points, l2_points)l0_points = self.fp1(l0_xyz, l1_xyz, None, l1_points)x = self.drop1(F.relu(self.bn1(self.conv1(l0_points))))x = self.conv2(x)x = F.log_softmax(x, dim=1)x = x.permute(0, 2, 1)return x, l4_pointsclass get_loss(nn.Module):def __init__(self):super(get_loss, self).__init__()self.gamma=2def forward(self, pred, target, trans_feat, weight):#pred: 模型预测的输出 target: 真实的标签或数据，用于计算损失total_loss = F.nll_loss(pred, target, weight=weight)return total_loss
if __name__ == '__main__':import torchmodel = get_model(13)xyz = torch.rand(6, 9, 2048)(model(xyz))
![[Flutter]WindowsOS中相关配置](https://img-blog.csdnimg.cn/direct/d3a457e5009a4477b2c643f4289f8cf5.png)
