🍨 本文为🔗365天深度学习训练营 中的学习记录博客
🍖 参考原作者:K同学啊|接辅导、项目定制
一、课题背景和开发环境
语言:Python3.8、Pytorch
📌本周任务:将yolov5s网络模型中的C3模块按照下图方式修改形成C2模块,并将C2模块插入第2层与第3层之间,且跑通yolov5。
💫任务提示:
– 提示1:需要修改./models/common.py、./models/yolo.py、./models/yolov5s.yaml文件
– 提示2:C2模块与C3模块是非常相似的两个模块,我们要插入C2到模型当中,只需要找到哪里有C3模块,然后在其附近加上C2即可。
二、代码解析
1、parse_model函数
这个函数用于将模型的模块拼接起来,搭建完整的网络模型。后续如果需要动模型框架的话,需要对这个函数做相应的改动。
def parse_model(d, ch): # model_dict, input_channels(3)
# Parse a YOLOv5 model.yaml dictionary
''' 用在上面DetectionModel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{dice: 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): 把所有层结构中的from不是-1的值记下,并排序[4,6,10,14,17,20,23]
'''
LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10} {'module':<40}{'arguments':<30}")
# 读取字典d中的anchors和parameters(nc,depth_multiple,width_multiple)
anchors, nc, gd, gw, act = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'], d.get('activation')
if act:
Conv.default_act = eval(act) # redefine default activation, i.e. Conv.default_act = nn.SiLU()
LOGGER.info(f"{colorstr('activation:')} {act}") # print
# 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 # number of anchors
# no: number of outputs 每一个predict head层的输出channel=anchors*(classes+5)=75(VOC)
no = na * (nc + 5) # number of outputs = anchors * (classes + 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不是-1的层结构序号
c2: 保存当前层的输出channel
'''
layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out
# from: 当前层输入来自哪些层
# number: 当前层数,初定
# module: 当前层类别
# args: 当前层类参数,初定
# 遍历backbone和head的每一层
for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # from, number, module, args
# 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 # eval strings
#
for j, a in enumerate(args):
with contextlib.suppress(NameError):
args[j] = eval(a) if isinstance(a, str) else a # eval strings
# --------------------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yolov5s: n*0.33,n: 当前模块的次数(间接控制深度)
n = n_ = max(round(n * gd), 1) if n > 1 else n # depth gain
if m in {
Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,
BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x}:
# c1: 当前层的输入channel数; c2: 当前层的输出channel数(初定); ch: 记录着所有层的输出channel数
c1, c2 = ch[f], args[0]
# 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: # if not output
# width gain 控制宽度,如yolov5s: c2*0.5; c2: 当前层的最终输出channel数(间接控制宽度)
c2 = make_divisible(c2 * gw, 8)
# 在初始args的基础上更新,加入当前层的输入channel并更新当前层
# [in_channels, out_channels, *args[1:]]
args = [c1, c2, *args[1:]]
# 如果当前层是BottleneckCSP/C3/C3TR/C3Ghost/C3x,则需要在args中加入Bottleneck的个数
# [in_channels, out_channels, Bottleneck个数, Bool(shortcut有无标记)]
if m in {BottleneckCSP, C3, C3TR, C3Ghost, C3x}:
args.insert(2, n) # number of repeats 在第二个位置插入Bottleneck的个数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)
# TODO: channel, gw, gd
elif m in Detect: # Detect/Segment(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)
if m is Segment:
args[3] = make_divisible(args[3] * gw, 8)
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
else: # Upsample
c2 = ch[f] # args不变
# -------------------------------------------------------------------------------------------
# m_: 得到当前层的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) # module
# 打印当前层结构的一些基本信息
t = str(m)[8:-2].replace('__main__.', '') # 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 # attach index, 'from' index, type, number params
LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f} {t:<40}{str(args):<30}') # print
# 把所有层结构中的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) # append to savelist
# 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)
params d: model_dict模型文件,字典形式{dice: 7}(yolov5s.yaml中的6个元素 + ch)
d['nc']
d['depth_multiple']
d['width_multiple']
d['anchors']
d['backbone']
d['head']
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): 把所有层结构中的from不是-1的值记下,并排序[4,6,10,14,17,20,23]
2.Detect类
Detect模块是用来构建Detect层的,将输入的feature map通过一个卷积操作和公式计算到我们想要的shape,为后面的计算损失率或者NMS做准备
class Detect(nn.Module):
# YOLOv5 Detect head for detection models
''' Detect模块是用来构建Detect层的
将输入的feature map通过一个卷积操作和公式计算到我们想要的shape,为后面的计算损失率或者NMS做准备
'''
stride = None # strides computed during build
dynamic = False # force grid reconstruction
export = False # export mode
def __init__(self, nc=80, anchors=(), ch=(), inplace=True): # detection layer
''' detection layer 相当于yolov3中的YOLO Layer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().__init__()
self.nc = nc # number of classes VOC: 20
self.no = nc + 5 # number of outputs per anchor VOC: 5(xywhc)+20(classes)=25
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.empty(0) for _ in range(self.nl)] # init grid {list: 3} tensor([0.])X3
self.anchor_grid = [torch.empty(0) for _ in range(self.nl)] # init anchor grid
''' 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;另一种不需要被更新,称为buffer
buffer的参数更新是在forward中,而optim.step只能更新nn.parameter参数
'''
self.register_buffer('anchors', torch.tensor(anchors).float().view(self.nl, -1, 2)) # shape(nl,na,2)
# output conv 对每个输出的feature map都要调用一次conv1 x 1
self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch) # output conv
# 一般都是True,默认不使用AWS,Inferentia加速
self.inplace = inplace # use inplace ops (e.g. slice assignment)
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+classes]
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+classes]
'''
z = [] # inference output
for i in range(self.nl): # 对3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 # x(bs,255,20,20) to x(bs,3,20,20,85)
# [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()
''' 构造网格
因为推理返回的不是归一化后的网络偏移量,需要加上网格的位置,得到最终的推理坐标,再送入NMS
所以这里构建网络就是为了记录每个grid的网格坐标,方便后面使用
'''
if not self.training: # inference
if self.dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:
self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)
if isinstance(self, Segment): # (boxes + masks)
xy, wh, conf, mask = x[i].split((2, 2, self.nc + 1, self.no - self.nc - 5), 4)
xy = (xy.sigmoid() * 2 + self.grid[i]) * self.stride[i] # xy
wh = (wh.sigmoid() * 2) ** 2 * self.anchor_grid[i] # wh
y = torch.cat((xy, wh, conf.sigmoid(), mask), 4)
else: # Detect (boxes only)
xy, wh, conf = x[i].sigmoid().split((2, 2, self.nc + 1), 4)
xy = (xy * 2 + self.grid[i]) * self.stride[i] # xy
wh = (wh * 2) ** 2 * self.anchor_grid[i] # wh
y = torch.cat((xy, wh, conf), 4)
# z是一个tensor list,有三个元素,分别是[1,19200,25] [1,4800,25] [1,1200,25]
z.append(y.view(bs, self.na * nx * ny, self.no))
return x if self.training else (torch.cat(z, 1),) if self.export else (torch.cat(z, 1), x)
def _make_grid(self, nx=20, ny=20, i=0, torch_1_10=check_version(torch.__version__, '1.10.0')):
''' 构造网格 '''
d = self.anchors[i].device
t = self.anchors[i].dtype
shape = 1, self.na, ny, nx, 2 # grid shape
y, x = torch.arange(ny, device=d, dtype=t), torch.arange(nx, device=d, dtype=t)
yv, xv = torch.meshgrid(y, x, indexing='ij') if torch_1_10 else torch.meshgrid(y, x) # torch>=0.7 compatibility
grid = torch.stack((xv, yv), 2).expand(shape) - 0.5 # add grid offset, i.e. y = 2.0 * x - 0.5
anchor_grid = (self.anchors[i] * self.stride[i]).view((1, self.na, 1, 1, 2)).expand(shape)
return grid, anchor_grid
3.Model类
class BaseModel(nn.Module):
# YOLOv5 base model
def forward(self, x, profile=False, visualize=False):
return self._forward_once(x, profile, visualize) # single-scale inference, train
def _forward_once(self, x, profile=False, visualize=False):
'''
:params x: 输入图像
:params profile: True 可以做一些性能评估
:params visualize: True 可以做一些特征可视化
:return train: 一个tensor,存放三个元素 [bs, anchor_num, grid_w, grid_h, xywh+c+classes]
inference: 0 [1,19200+4800+1200,25]=[bs,anchor_num*grid_w*grid_h,xywh+c+classes]
'''
# 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 = [], [] # outputs
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 parameters
if m.f != -1: # if not from previous layer m.f=当前层的输入来自哪一层的输出,-1表示上一层
# 这里需要做4个Concat操作和一个Detect操作
# Concat: 如m.f=[-1,6] x就有两个元素,一个是上一层的输出,一个是index=6的层的输出,再送到x=m(x)做Concat操作
# Detect: 如m.f=[17, 20, 23] x就有三个元素,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:
self._profile_one_layer(m, x, dt)
x = m(x) # run 正向推理
# 存放着self.save的每一层的输出,因为后面需要用来做Concat等操作,不在self.save层的输出就为None
y.append(x if m.i in self.save else None) # save output
# 特征可视化,可以自己改动想要那层的特征进行可视化
if visualize:
feature_visualization(x, m.type, m.i, save_dir=visualize)
return x
def _profile_one_layer(self, m, x, dt):
c = m == self.model[-1] # is final layer, copy input as inplace fix
o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPs
t = time_sync()
for _ in range(10):
m(x.copy() if c else x)
dt.append((time_sync() - 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}')
if c:
LOGGER.info(f"{sum(dt):10.2f} {'-':>10s} {'-':>10s} Total")
def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers
''' 用在detect.py、val.py中
fuse model Conv2d() + BatchNorm2d() layers
调用torch_utils.py中的fuse_conv_and_bn函数和common.py中的forward_fuse函数
'''
LOGGER.info('Fusing layers... ') # 日志
for m in self.model.modules(): # 遍历每一层结构
# 如果当前层是卷积层Conv且有BN结构,那么就调用fuse_conv_and_bn函数将Conv和BN进行融合,加速推理
if isinstance(m, (Conv, DWConv)) and hasattr(m, 'bn'):
m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv 融合
delattr(m, 'bn') # remove batchnorm 移除BN
m.forward = m.forward_fuse # update forward 更新前向传播(反向传播不用管,因为这个过程只用再推理阶段)
self.info() # 打印Conv+BN融合后的模型信息
return self
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 _apply(self, fn):
# Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers
self = super()._apply(fn)
m = self.model[-1] # Detect()
if isinstance(m, (Detect, Segment)):
m.stride = fn(m.stride)
m.grid = list(map(fn, m.grid))
if isinstance(m.anchor_grid, list):
m.anchor_grid = list(map(fn, m.anchor_grid))
return self
class DetectionModel(BaseModel):
# YOLOv5 detection model
def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None): # model, input channels, number of classes
'''
: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
:params anchors: 一般是None
'''
super().__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='ascii', errors='ignore') as f: # encoding='utf-8'
self.yaml = yaml.safe_load(f) # model dict
# Define model
ch = self.yaml['ch'] = self.yaml.get('ch', ch) # input channels ch=3
# 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anchors,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]) # model, savelist
# default class names ['0','1','2',...,'19']
self.names = [str(i) for i in range(self.yaml['nc'])] # default names
# 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)
# Build strides, anchors
# 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的尺寸
m = self.model[-1] # Detect()
if isinstance(m, (Detect, Segment)):
s = 256 # 2x min stride
m.inplace = self.inplace
forward = lambda x: self.forward(x)[0] if isinstance(m, Segment) else self.forward(x)
# 计算三个feature map的anchor大小,如[10,13]/8 -> [1.25,1.625]
m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))]) # forward
# 检查anchor顺序与stride顺序是否一致
check_anchor_order(m)
m.anchors /= m.stride.view(-1, 1, 1)
self.stride = m.stride
self._initialize_biases() # only run once 初始化偏置
# 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, visualize=False):
# 是否在测试时也使用数据增强 Test Time Augmentation(TTA)
if augment:
return self._forward_augment(x) # augmented inference, None 上下flip/左右flip
# 默认执行,正常前向推理
return self._forward_once(x, profile, visualize) # single-scale inference, train
def _forward_augment(self, x):
''' TTA Test Time Augmentation '''
img_size = x.shape[-2:] # height, width
s = [1, 0.83, 0.67] # scales
f = [None, 3, None] # flips (2-ud上下, 3-lr左右)
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)
y = self._clip_augmented(y) # clip augmented tails
return torch.cat(y, 1), None # augmented inference, train
def _descale_pred(self, p, flips, scale, img_size):
# de-scale predictions following augmented inference (inverse operation)
''' 用在上面的__init__函数上
将推理结果恢复到原图图片尺寸上 TTA中用到
:params p: 推理结果
:params flips: 翻转标记(2-ud上下, 3-lr左右)
:params scale: 图片缩放比例
:params img_size: 原图图片尺寸
'''
# 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True,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 _clip_augmented(self, y):
# Clip YOLOv5 augmented inference tails
nl = self.model[-1].nl # number of detection layers (P3-P5)
g = sum(4 ** x for x in range(nl)) # grid points
e = 1 # exclude layer count
i = (y[0].shape[1] // g) * sum(4 ** x for x in range(e)) # indices
y[0] = y[0][:, :-i] # large
i = (y[-1].shape[1] // g) * sum(4 ** (nl - 1 - x) for x in range(e)) # indices
y[-1] = y[-1][:, i:] # small
return y
def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency
''' 用在上面的__init__函数上 '''
# 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[:, 5:5 + m.nc] += math.log(0.6 / (m.nc - 0.99999)) if cf is None else torch.log(cf / cf.sum()) # cls
mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
Model = DetectionModel # retain YOLOv5 'Model' class for backwards compatibility
三、调整模型
在yolov5s.yaml原第2层和原第3层之间插入C2模块
# YOLOv5 v6.0 backbone
backbone:
# [from, number, module, args]
[[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2
[-1, 1, Conv, [128, 3, 2]], # 1-P2/4
[-1, 3, C3, [128]],
[-1, 3, C2, [128]], # 在原第2层和原第3层之间插入C2模块
[-1, 1, Conv, [256, 3, 2]], # 3-P3/8
[-1, 6, C3, [256]],
[-1, 1, Conv, [512, 3, 2]], # 5-P4/16
[-1, 9, C3, [512]],
[-1, 1, Conv, [1024, 3, 2]], # 7-P5/32
[-1, 3, C3, [1024]],
[-1, 1, SPPF, [1024, 5]], # 9
]
common.py 加C2模块
在parse_model
中增加对C2的解析