YOLOv7-u6分支的实现是基于Yolov5和Yolov6进行的。并在此基础上开发了Anchor-Free方法。所有安装、数据准备和使用与Yolov5相同，大家可以酌情尝试，如果电费不要钱，那就不要犹豫了！！！
先看原始的YOLOv7的精度
当时原始版本就是无敌的存在，YOLOv7的base版本就有51.2的精度了！！！
再看Anchor-Free版本YOLOv7的精度
再看原作复现的Anchor-Free版本，相对于原始版本的51.2的精度，分别提升了1.1个点和1.4个点（使用了albumentation数据增强），可以看出还是很给力的结构。
架构改进部分
其实，关于复现的YOLOv7-u6（Anchor-Free），Backbone和Neck部分是没有发生变化的，下面看一下Head部分的变化。
1、YOLOv7的Anchor-Base Head
通过下图的YAML知道，YOLOv7的head使用了重参结构，并且也加入了隐藏知识Trick的加入。
2、YOLOv7的Anchor-Free Head
去除了RepConv卷积，使用了最为基本的Conv模块，同时检测头换为了YOLOv6的Head形式，同时加入了IDetect的隐藏知识Implicit层思想。
3、IV6Detect的实现如下
class IV6Detect(nn.Module): dynamic = False # force grid reconstruction export = False # export mode shape = None anchors = torch.empty(0) # init strides = torch.empty(0) # init def __init__(self, nc=80, ch=(), inplace=True): # detection layer super().__init__() self.nc = nc # number of classes self.nl = len(ch) # number of detection layers self.reg_max = 16 self.no = nc + self.reg_max * 4 # number of outputs per anchor self.inplace = inplace # use inplace ops (e.g. slice assignment) self.stride = torch.zeros(self.nl) # strides computed during build c2, c3 = max(ch[0] // 4, 16), max(ch[0], self.no - 4) # channels self.cv2 = nn.ModuleList( nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch) self.cv3 = nn.ModuleList( nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch) # DFL层 self.dfl = DFL(self.reg_max) # Implicit层 self.ia2 = nn.ModuleList(ImplicitA(x) for x in ch) self.ia3 = nn.ModuleList(ImplicitA(x) for x in ch) self.im2 = nn.ModuleList(ImplicitM(4 * self.reg_max) for _ in ch) self.im3 = nn.ModuleList(ImplicitM(self.nc) for _ in ch) def forward(self, x): shape = x[0].shape # BCHW for i in range(self.nl): x[i] = torch.cat((self.im2[i](self.cv2[i](self.ia2[i](x[i]))), self.im3[i](self.cv3[i](self.ia3[i](x[i])))), 1) box, cls = torch.cat([xi.view(shape[0], self.no, -1) for xi in x], 2).split((self.reg_max * 4, self.nc), 1) if self.training: return x, box, cls elif self.dynamic or self.shape != shape: self.anchors, self.strides = (x.transpose(0, 1) for x in make_anchors(x, self.stride, 0.5)) self.shape = shape dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.strides y = torch.cat((dbox, cls.sigmoid()), 1) return y if self.export else (y, (x, box, cls)) def bias_init(self): m = self # self.model[-1] # Detect() module for a, b, s in zip(m.cv2, m.cv3, m.stride): # from a[-1].bias.data[:] = 1.0 # box b[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)
关于损失函数与样本匹配的穿搭
一句话吧，其实就是YOLOv8本来的样子，也可能YOLOv8是原来YOLOv7-u6本来的样子。使用了TaskAligned Assigner，BCE Loss、CIOU Loss以及DFL Loss。可以说是标准搭配了！！！
class ComputeLoss: def __init__(self, model, use_dfl=True): device = next(model.parameters()).device # get model device h = model.hyp # hyperparameters # Define criteria # 分类损失 BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h["cls_pw"]], device=device), reduction='none') # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3 self.cp, self.cn = smooth_BCE(eps=h.get("label_smoothing", 0.0)) # positive, negative BCE targets # Focal loss g = h["fl_gamma"] # focal loss gamma if g > 0: BCEcls = FocalLoss(BCEcls, g) m = de_parallel(model).model[-1] # Detect() module self.balance = {3: [4.0, 1.0, 0.4]}.get(m.nl, [4.0, 1.0, 0.25, 0.06, 0.02]) # P3-P7 self.BCEcls = BCEcls self.hyp = h self.stride = m.stride # model strides self.nc = m.nc # number of classes self.nl = m.nl # number of layers self.device = device # 正负样本匹配 self.assigner = TaskAlignedAssigner(topk=int(os.getenv('YOLOM', 10)), num_classes=self.nc, alpha=float(os.getenv('YOLOA', 0.5)), beta=float(os.getenv('YOLOB', 6.0))) # 回归损失函数 self.bbox_loss = BboxLoss(m.reg_max - 1, use_dfl=use_dfl).to(device) self.proj = torch.arange(m.reg_max).float().to(device) # / 120.0 self.use_dfl = use_dfl
参考
[1].https://github.com/WongKinYiu/yolov7/tree/u6.
