内容
- 前言
- 0、导入需要的包和基本配置
- 1、parse_model
- 2、Detect
- 3、Model
- 总结
前言
源码：YOLOv5源码.
导航:【YOLOV5-5.x 源码讲解】整体项目文件导航.
注释版全部项目文件已上传至GitHub:yolov5-5.x-annotations.
0、导入需要的包和基本配置
import argparse # 解析命令行参数模块
import logging # 日志模块
import sys # sys系统模块 包含了与Python解释器和它的环境有关的函数
from copy import deepcopy # 数据拷贝模块 深拷贝
from pathlib import Path # Path将str转换为Path对象 使字符串路径易于操作的模块
FILE = Path(__file__).absolute() # FILE = WindowsPath 'F:\yolo_v5\yolov5-U\modles\yolo.py'
# 将'F:/yolo_v5/yolov5-U'加入系统的环境变量 该脚本结束后失效
sys.path.append(FILE.parents[1].as_posix()) # add yolov5/ to path
from models.common import *
from models.experimental import *
from utils.autoanchor import check_anchor_order
from utils.general import make_pisible, check_file, set_logging
from utils.plots import feature_visualization
from utils.torch_utils import time_synchronized, fuse_conv_and_bn, model_info, \
scale_img, initialize_weights, select_device, copy_attr
# 导入thop包 用于计算FLOPs
try:
import thop # for FLOPs computation
except ImportError:
thop = None
# 初始化日志
logger = logging.getLogger(__name__)
1、parse_model
parse_model模块代码
def parse_model(d, ch): # model_dict, input_channels(3)
"""用在上面Model模块中
解析模型文件(字典形式)，并搭建网络结构
这个函数其实主要做的就是: 更新当前层的args（参数）,计算c2（当前层的输出channel） =>
使用当前层的参数搭建当前层 =>
生成 layers + save
:params d: model_dict 模型文件 字典形式 {dict:7} yolov5s.yaml中的6个元素 + ch
:params ch: 记录模型每一层的输出channel 初始ch=[3] 后面会删除
:return nn.Sequential(*layers): 网络的每一层的层结构
:return sorted(save): 把所有层结构中from不是-1的值记下 并排序 [4, 6, 10, 14, 17, 20, 23]
"""
logger.info('\n%3s%18s%3s%10s %-40s%-30s' % ('', 'from', 'n', 'params', 'module', 'arguments'))
# 读取d字典中的anchors和parameters(nc、depth_multiple、width_multiple)
anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
# na: number of anchors 每一个predict head上的anchor数 = 3
na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors
# no: number of outputs 每一个predict head层的输出channel = anchors * (classes + 5) = 75(VOC)
no = na * (nc + 5)
# 开始搭建网络
# layers: 保存每一层的层结构
# save: 记录下所有层结构中from中不是-1的层结构序号
# c2: 保存当前层的输出channel
layers, save, c2 = [], [], ch[-1]
# from(当前层输入来自哪些层), number(当前层次数 初定), module(当前层类别), args(当前层类参数 初定)
for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # 遍历backbone和head的每一层
# eval(string) 得到当前层的真实类名 例如: m= Focus -> <class 'models.common.Focus'>
m = eval(m) if isinstance(m, str) else m
# 没什么用
for j, a in enumerate(args):
try:
args[j] = eval(a) if isinstance(a, str) else a # eval strings
except:
pass
# ------------------- 更新当前层的args（参数）,计算c2（当前层的输出channel） -------------------
# depth gain 控制深度 如v5s: n*0.33 n: 当前模块的次数(间接控制深度)
n = max(round(n * gd), 1) if n > 1 else n
if m in [Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, DWConv, MixConv2d,
Focus, CrossConv, BottleneckCSP, C3, C3TR, CBAM]:
# c1: 当前层的输入的channel数 c2: 当前层的输出的channel数(初定) ch: 记录着所有层的输出channel
c1, c2 = ch[f], args[0]
# if not output no=75 只有最后一层c2=no 最后一层不用控制宽度，输出channel必须是no
if c2 != no:
# width gain 控制宽度 如v5s: c2*0.5 c2: 当前层的最终输出的channel数(间接控制宽度)
c2 = make_pisible(c2 * gw, 8)
# 在初始arg的基础上更新 加入当前层的输入channel并更新当前层
# [in_channel, out_channel, *args[1:]]
args = [c1, c2, *args[1:]]
# 如果当前层是BottleneckCSP/C3/C3TR, 则需要在args中加入bottleneck的个数
# [in_channel, out_channel, Bottleneck的个数n, bool(True表示有shortcut 默认，反之无)]
if m in [BottleneckCSP, C3, C3TR]:
args.insert(2, n) # 在第二个位置插入bottleneck个数n
n = 1 # 恢复默认值1
elif m is nn.BatchNorm2d:
# BN层只需要返回上一层的输出channel
args = [ch[f]]
elif m is Concat:
# Concat层则将f中所有的输出累加得到这层的输出channel
c2 = sum([ch[x] for x in f])
elif m is Detect: # Detect（YOLO Layer）层
# 在args中加入三个Detect层的输出channel
args.append([ch[x] for x in f])
if isinstance(args[1], int): # number of anchors 几乎不执行
args[1] = [list(range(args[1] * 2))] * len(f)
elif m is Contract: # 不怎么用
c2 = ch[f] * args[0] ** 2
elif m is Expand: # 不怎么用
c2 = ch[f] // args[0] ** 2
elif m is SELayer: # 加入SE模块
channel, re = args[0], args[1]
channel = make_pisible(channel * gw, 8) if channel != no else channel
args = [channel, re]
else:
# Upsample
c2 = ch[f] # args不变
# -----------------------------------------------------------------------------------
# m_: 得到当前层module 如果n>1就创建多个m(当前层结构), 如果n=1就创建一个m
m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n > 1 else m(*args)
# 打印当前层结构的一些基本信息
t = str(m)[8:-2].replace('__main__.', '') # t = module type 'modules.common.Focus'
np = sum([x.numel() for x in m_.parameters()]) # number params 计算这一层的参数量
m_.i, m_.f, m_.type, m_.np = i, f, t, np # index, 'from' index, number, type, number params
logger.info('%3s%18s%3s%10.0f %-40s%-30s' % (i, f, n, np, t, args)) # print
# append to savelist 把所有层结构中from不是-1的值记下 [6, 4, 14, 10, 17, 20, 23]
save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)
# 将当前层结构module加入layers中
layers.append(m_)
if i == 0:
ch = [] # 去除输入channel [3]
# 把当前层的输出channel数加入ch
ch.append(c2)
return nn.Sequential(*layers), sorted(save)
在Model模块的__init__函数中调用：
2、Detect
Detect模块代码：
class Detect(nn.Module):
"""Detect模块是用来构建Detect层的，将输入feature map 通过一个卷积操作和公式计算到我们想要的shape, 为后面的计算损失或者NMS作准备"""
stride = None # strides computed during build
onnx_dynamic = False # ONNX export parameter
def __init__(self, nc=80, anchors=(), ch=(), inplace=True):
"""
detection layer 相当于yolov3中的YOLOLayer层
:params nc: number of classes
:params anchors: 传入3个feature map上的所有anchor的大小（P3、P4、P5）
:params ch: [128, 256, 512] 3个输出feature map的channel
"""
super(Detect, self).__init__()
self.nc = nc # number of classes VOC: 20
self.no = nc + 5 # number of outputs per anchor VOC: 5+20=25 xywhc+20classes
self.nl = len(anchors) # number of detection layers Detect的个数 3
self.na = len(anchors[0]) // 2 # number of anchors 每个feature map的anchor个数 3
self.grid = [torch.zeros(1)] * self.nl # init grid {list: 3} tensor([0.]) X 3
# a=[3, 3, 2] anchors以[w, h]对的形式存储 3个feature map 每个feature map上有三个anchor（w,h）
a = torch.tensor(anchors).float().view(self.nl, -1, 2)
# register_buffer
# 模型中需要保存的参数一般有两种：一种是反向传播需要被optimizer更新的，称为parameter; 另一种不要被更新称为buffer
# buffer的参数更新是在forward中，而optim.step只能更新nn.parameter类型的参数
# shape(nl,na,2)
self.register_buffer('anchors', a)
# shape(nl,1,na,1,1,2)
self.register_buffer('anchor_grid', a.clone().view(self.nl, 1, -1, 1, 1, 2))
# output conv 对每个输出的feature map都要调用一次conv1x1
self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch)
# use in-place ops (e.g. slice assignment) 一般都是True 默认不使用AWS Inferentia加速
self.inplace = inplace
def forward(self, x):
"""
:return train: 一个tensor list 存放三个元素 [bs, anchor_num, grid_w, grid_h, xywh+c+20classes]
分别是 [1, 3, 80, 80, 25] [1, 3, 40, 40, 25] [1, 3, 20, 20, 25]
inference: 0 [1, 19200+4800+1200, 25] = [bs, anchor_num*grid_w*grid_h, xywh+c+20classes]
1 一个tensor list 存放三个元素 [bs, anchor_num, grid_w, grid_h, xywh+c+20classes]
[1, 3, 80, 80, 25] [1, 3, 40, 40, 25] [1, 3, 20, 20, 25]
"""
# x = x.copy() # for profiling
z = [] # inference output
for i in range(self.nl): # 对三个feature map分别进行处理
x[i] = self.m[i](x[i]) # conv xi[bs, 128/256/512, 80, 80] to [bs, 75, 80, 80]
bs, _, ny, nx = x[i].shape
# [bs, 75, 80, 80] to [1, 3, 25, 80, 80] to [1, 3, 80, 80, 25]
x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
# inference
if not self.training:
# 构造网格
# 因为推理返回的不是归一化后的网格偏移量 需要再加上网格的位置 得到最终的推理坐标 再送入nms
# 所以这里构建网格就是为了纪律每个grid的网格坐标 方面后面使用
if self.grid[i].shape[2:4] != x[i].shape[2:4] or self.onnx_dynamic:
self.grid[i] = self._make_grid(nx, ny).to(x[i].device)
y = x[i].sigmoid()
if self.inplace:
# 默认执行 不使用AWS Inferentia
# 这里的公式和yolov3、v4中使用的不一样 是yolov5作者自己用的 效果更好
y[..., 0:2] = (y[..., 0:2] * 2. - 0.5 + self.grid[i]) * self.stride[i] # xy
# y[..., 2:4] = torch.exp(y[..., 2:4]) * self.anchor_wh # wh yolo method
y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh power method
else: # for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953
xy = (y[..., 0:2] * 2. - 0.5 + self.grid[i]) * self.stride[i] # xy
wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i].view(1, self.na, 1, 1, 2) # wh
y = torch.cat((xy, wh, y[..., 4:]), -1)
# z是一个tensor list 三个元素 分别是[1, 19200, 25] [1, 4800, 25] [1, 1200, 25]
z.append(y.view(bs, -1, self.no))
return x if self.training else (torch.cat(z, 1), x)
@staticmethod
def _make_grid(nx=20, ny=20):
"""
构造网格
"""
yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)])
return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float()
__init__函数在parse_model函数中调用：
forward函数在Model类的forward_once中调用：
3、Model
Model模块代码:
class Model(nn.Module):
def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None):
"""
:params cfg:模型配置文件
:params ch: input img channels 一般是3 RGB文件
:params nc: number of classes 数据集的类别个数
:anchors: 一般是None
"""
super(Model, self).__init__()
if isinstance(cfg, dict):
self.yaml = cfg # model dict
else:
# is *.yaml 一般执行这里
import yaml # for torch hub
self.yaml_file = Path(cfg).name # cfg file name = yolov5s.yaml
# 如果配置文件中有中文，打开时要加encoding参数
with open(cfg, encoding='utf-8') as f:
# model dict 取到配置文件中每条的信息（没有注释内容）
self.yaml = yaml.safe_load(f)
# input channels ch=3
ch = self.yaml['ch'] = self.yaml.get('ch', ch)
# 设置类别数 一般不执行, 因为nc=self.yaml['nc']恒成立
if nc and nc != self.yaml['nc']:
logger.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")
self.yaml['nc'] = nc # override yaml value
# 重写anchor，一般不执行, 因为传进来的anchors一般都是None
if anchors:
logger.info(f'Overriding model.yaml anchors with anchors={anchors}')
self.yaml['anchors'] = round(anchors) # override yaml value
# 创建网络模型
# self.model: 初始化的整个网络模型(包括Detect层结构)
# self.save: 所有层结构中from不等于-1的序号，并排好序 [4, 6, 10, 14, 17, 20, 23]
self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch])
# default class names ['0', '1', '2',..., '19']
self.names = [str(i) for i in range(self.yaml['nc'])]
# self.inplace=True 默认True 不使用加速推理
# AWS Inferentia Inplace compatiability
# https://github.com/ultralytics/yolov5/pull/2953
self.inplace = self.yaml.get('inplace', True)
# logger.info([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))])
# 获取Detect模块的stride(相对输入图像的下采样率)和anchors在当前Detect输出的feature map的尺度
m = self.model[-1] # Detect()
if isinstance(m, Detect):
s = 256 # 2x min stride
m.inplace = self.inplace
# 计算三个feature map下采样的倍率 [8, 16, 32]
m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward
# 求出相对当前feature map的anchor大小 如[10, 13]/8 -> [1.25, 1.625]
m.anchors /= m.stride.view(-1, 1, 1)
# 检查anchor顺序与stride顺序是否一致
check_anchor_order(m)
self.stride = m.stride
self._initialize_biases() # only run once 初始化偏置
# logger.info('Strides: %s' % m.stride.tolist())
# Init weights, biases
initialize_weights(self) # 调用torch_utils.py下initialize_weights初始化模型权重
self.info() # 打印模型信息
logger.info('')
def forward(self, x, augment=False, profile=False):
# augmented inference, None 上下flip/左右flip
# 是否在测试时也使用数据增强 Test Time Augmentation(TTA)
if augment:
return self.forward_augment(x)
else:
# 默认执行 正常前向推理
# single-scale inference, train
return self.forward_once(x, profile)
def forward_augment(self, x):
"""
TTA Test Time Augmentation
"""
img_size = x.shape[-2:] # height, width
s = [1, 0.83, 0.67] # scales ratio
f = [None, 3, None] # flips (2-ud上下flip, 3-lr左右flip)
y = [] # outputs
for si, fi in zip(s, f):
# scale_img缩放图片尺寸
xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max()))
yi = self.forward_once(xi)[0] # forward
# cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1]) # save
# _descale_pred将推理结果恢复到相对原图图片尺寸
yi = self._descale_pred(yi, fi, si, img_size)
y.append(yi)
return torch.cat(y, 1), None # augmented inference, train
def forward_once(self, x, profile=False, feature_vis=False):
"""
:params x: 输入图像
:params profile: True 可以做一些性能评估
:params feature_vis: True 可以做一些特征可视化
:return train: 一个tensor list 存放三个元素 [bs, anchor_num, grid_w, grid_h, xywh+c+20classes]
分别是 [1, 3, 80, 80, 25] [1, 3, 40, 40, 25] [1, 3, 20, 20, 25]
inference: 0 [1, 19200+4800+1200, 25] = [bs, anchor_num*grid_w*grid_h, xywh+c+20classes]
1 一个tensor list 存放三个元素 [bs, anchor_num, grid_w, grid_h, xywh+c+20classes]
[1, 3, 80, 80, 25] [1, 3, 40, 40, 25] [1, 3, 20, 20, 25]
"""
# y: 存放着self.save=True的每一层的输出，因为后面的层结构concat等操作要用到
# dt: 在profile中做性能评估时使用
y, dt = [], []
for m in self.model:
# 前向推理每一层结构 m.i=index m.f=from m.type=类名 m.np=number of params
# if not from previous layer m.f=当前层的输入来自哪一层的输出 s的m.f都是-1
if m.f != -1:
# 这里需要做4个concat操作和1个Detect操作
# concat操作如m.f=[-1, 6] x就有两个元素,一个是上一层的输出,另一个是index=6的层的输出 再送到x=m(x)做concat操作
# Detect操作m.f=[17, 20, 23] x有三个元素,分别存放第17层第20层第23层的输出 再送到x=m(x)做Detect的forward
x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers
# 打印日志信息 FLOPs time等
if profile:
o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPs
t = time_synchronized()
for _ in range(10):
_ = m(x)
dt.append((time_synchronized() - t) * 100)
if m == self.model[0]:
logger.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s} {'module'}")
logger.info(f'{dt[-1]:10.2f} {o:10.2f} {m.np:10.0f} {m.type}')
x = m(x) # run正向推理 执行每一层的forward函数(除Concat和Detect操作)
# 存放着self.save的每一层的输出，因为后面需要用来作concat等操作要用到 不在self.save层的输出就为None
y.append(x if m.i in self.save else None)
# 特征可视化 可以自己改动想要哪层的特征进行可视化
if feature_vis and m.type == 'models.common.SPP':
feature_visualization(x, m.type, m.i)
# 打印日志信息 前向推理时间
if profile:
logger.info('%.1fms total' % sum(dt))
return x
def _initialize_biases(self, cf=None):
"""用在上面的__init__函数上
initialize biases into Detect(), cf is class frequency
https://arxiv.org/abs/1708.02002 section 3.3
"""
# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.
m = self.model[-1] # Detect() module
for mi, s in zip(m.m, m.stride): # from
b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85)
b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image)
b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum()) # cls
mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
def info(self, verbose=False, img_size=640): # print model information
"""用在上面的__init__函数上
调用torch_utils.py下model_info函数打印模型信息
"""
model_info(self, verbose, img_size)
def _descale_pred(self, p, flips, scale, img_size):
"""用在上面的__init__函数上
将推理结果恢复到原图图片尺寸 Test Time Augmentation(TTA)中用到
de-scale predictions following augmented inference (inverse operation)
:params p: 推理结果
:params flips:
:params scale:
:params img_size:
"""
# 不同的方式前向推理使用公式不同 具体可看Detect函数
if self.inplace: # 默认执行 不使用AWS Inferentia
p[..., :4] /= scale # de-scale
if flips == 2:
p[..., 1] = img_size[0] - p[..., 1] # de-flip ud
elif flips == 3:
p[..., 0] = img_size[1] - p[..., 0] # de-flip lr
else:
x, y, wh = p[..., 0:1] / scale, p[..., 1:2] / scale, p[..., 2:4] / scale # de-scale
if flips == 2:
y = img_size[0] - y # de-flip ud
elif flips == 3:
x = img_size[1] - x # de-flip lr
p = torch.cat((x, y, wh, p[..., 4:]), -1)
return p
def _print_biases(self):
"""
打印模型中最后Detect层的偏置bias信息(也可以任选哪些层bias信息)
"""
m = self.model[-1] # Detect() module
for mi in m.m: # from
b = mi.bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85)
logger.info(
('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))
def _print_weights(self):
"""
打印模型中Bottleneck层的权重参数weights信息(也可以任选哪些层weights信息)
"""
for m in self.model.modules():
if type(m) is Bottleneck:
logger.info('%10.3g' % (m.w.detach().sigmoid() * 2)) # shortcut weights
def fuse(self):
"""用在detect.py、val.py
fuse model Conv2d() + BatchNorm2d() layers
调用torch_utils.py中的fuse_conv_and_bn函数和common.py中Conv模块的fuseforward函数
"""
logger.info('Fusing layers... ') # 日志
# 遍历每一层结构
for m in self.model.modules():
# 如果当前层是卷积层Conv且有bn结构, 那么就调用fuse_conv_and_bn函数讲conv和bn进行融合, 加速推理
if type(m) is Conv and hasattr(m, 'bn'):
m.conv = fuse_conv_and_bn(m.conv, m.bn) # 融合 update conv
delattr(m, 'bn') # 移除bn remove batchnorm
m.forward = m.fuseforward # 更新前向传播 update forward (反向传播不用管, 因为这种推理只用在推理阶段)
self.info() # 打印conv+bn融合后的模型信息
return self
def nms(self, mode=True):
"""
add or remove NMS module
可以自选是否扩展model 增加模型nms功能 直接调用common.py中的NMS模块
一般是用不到的 前向推理结束直接掉用non_max_suppression函数即可
"""
present = type(self.model[-1]) is NMS # last layer is NMS
if mode and not present:
logger.info('Adding NMS... ')
m = NMS() # module
m.f = -1 # from
m.i = self.model[-1].i + 1 # index
self.model.add_module(name='%s' % m.i, module=m) # add nms module to model
self.eval() # nms 开启模型验证模式
elif not mode and present:
logger.info('Removing NMS... ')
self.model = self.model[:-1] # remove nms from model
return self
def autoshape(self):
"""
add AutoShape module 直接调用common.py中的AutoShape模块 也是一个扩展模型功能的模块
"""
logger.info('Adding AutoShape... ')
# wrap model 扩展模型功能 此时模型包含前处理、推理、后处理的模块(预处理 + 推理 + nms)
m = AutoShape(self)
copy_attr(m, self, include=('yaml', 'nc', 'hyp', 'names', 'stride'), exclude=()) # copy attributes
return m
总结
– 2021.08.23 2021.08.23
版权声明：本文为博主满船清梦压星河HK原创文章，版权归属原作者，如果侵权，请联系我们删除！
原文链接：https://blog.csdn.net/qq_38253797/article/details/119869762