diff --git a/models/yolo.py b/models/yolo.py index 13498ac..1198e9c 100644 --- a/models/yolo.py +++ b/models/yolo.py @@ -125,6 +125,80 @@ class Detect(nn.Module): 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 + +# 新增 +class DetectDecoupledLite(nn.Module): + stride = None + dynamic = False + export = False + + def __init__(self, nc=80, anchors=(), ch=(), inplace=True): + super().__init__() + self.nc = nc + self.no = nc + 5 + self.nl = len(anchors) + self.na = len(anchors[0]) // 2 + self.grid = [torch.empty(0) for _ in range(self.nl)] + self.anchor_grid = [torch.empty(0) for _ in range(self.nl)] + self.register_buffer("anchors", torch.tensor(anchors).float().view(self.nl, -1, 2)) + self.inplace = inplace + + head_c = 128 + + # stem + self.stems = nn.ModuleList(Conv(c, head_c, 1, 1) for c in ch) + + # reg + obj shared trunk + self.regobj_convs = nn.ModuleList(Conv(head_c, head_c, 3, 1) for _ in ch) + + # cls trunk + self.cls_convs = nn.ModuleList(Conv(head_c, head_c, 3, 1) for _ in ch) + + # predictors + self.box_preds = nn.ModuleList(nn.Conv2d(head_c, self.na * 4, 1) for _ in ch) + self.obj_preds = nn.ModuleList(nn.Conv2d(head_c, self.na * 1, 1) for _ in ch) + self.cls_preds = nn.ModuleList(nn.Conv2d(head_c, self.na * self.nc, 1) for _ in ch) + + def forward(self, x): + z = [] + for i in range(self.nl): + xi = self.stems[i](x[i]) + + regobj_feat = self.regobj_convs[i](xi) + cls_feat = self.cls_convs[i](xi) + + box = self.box_preds[i](regobj_feat) + obj = self.obj_preds[i](regobj_feat) + cls = self.cls_preds[i](cls_feat) + + # [bs, na*4 + na*1 + na*nc, ny, nx] + out = torch.cat((box, obj, cls), 1) + + bs, _, ny, nx = out.shape + out = out.view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous() + x[i] = out + + if not self.training: + 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) + + xy, wh, conf = x[i].sigmoid().split((2, 2, self.nc + 1), 4) + xy = (xy * 2 + self.grid[i]) * self.stride[i] + wh = (wh * 2) ** 2 * self.anchor_grid[i] + y = torch.cat((xy, wh, conf), 4) + 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")): + d = self.anchors[i].device + t = self.anchors[i].dtype + shape = 1, self.na, ny, nx, 2 + 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) + grid = torch.stack((xv, yv), 2).expand(shape) - 0.5 + anchor_grid = (self.anchors[i] * self.stride[i]).view((1, self.na, 1, 1, 2)).expand(shape) + return grid, anchor_grid class Segment(Detect): @@ -207,7 +281,8 @@ class BaseModel(nn.Module): """ self = super()._apply(fn) m = self.model[-1] # Detect() - if isinstance(m, (Detect, Segment)): + # if isinstance(m, (Detect, Segment)): + if isinstance(m, (Detect, DetectDecoupledLite, Segment)): m.stride = fn(m.stride) m.grid = list(map(fn, m.grid)) if isinstance(m.anchor_grid, list): @@ -244,7 +319,8 @@ class DetectionModel(BaseModel): # Build strides, anchors m = self.model[-1] # Detect() - if isinstance(m, (Detect, Segment)): + # if isinstance(m, (Detect, Segment)): + if isinstance(m, (Detect, DetectDecoupledLite, Segment)): def _forward(x): """Passes the input 'x' through the model and returns the processed output.""" @@ -256,12 +332,36 @@ class DetectionModel(BaseModel): check_anchor_order(m) m.anchors /= m.stride.view(-1, 1, 1) self.stride = m.stride - self._initialize_biases() # only run once + # self._initialize_biases() # only run once + if isinstance(m, Detect): + self._initialize_biases() + elif isinstance(m, DetectDecoupledLite): + self._initialize_decoupled_biases() # Init weights, biases initialize_weights(self) self.info() LOGGER.info("") + + #新增 + def _initialize_decoupled_biases(self, cf=None): + """Initialize biases for DetectDecoupledLite.""" + m = self.model[-1] # DetectDecoupledLite + + # objectness bias + for mi, s in zip(m.obj_preds, m.stride): + b = mi.bias.view(m.na, -1) # (na, 1) + b.data += math.log(8 / (640 / s) ** 2) + mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) + + # classification bias + for mi in m.cls_preds: + b = mi.bias.view(m.na, -1) # (na, nc) + if cf is None: + b.data += math.log(0.6 / (m.nc - 0.99999)) + else: + b.data += torch.log(cf / cf.sum()) + mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) def forward(self, x, augment=False, profile=False, visualize=False): """Performs single-scale or augmented inference and may include profiling or visualization.""" @@ -434,7 +534,8 @@ def parse_model(d, ch): elif m is Concat: c2 = sum(ch[x] for x in f) # TODO: channel, gw, gd - elif m in {Detect, Segment}: + # elif m in {Detect, Segment}: + elif m in {Detect, DetectDecoupledLite, Segment}: args.append([ch[x] for x in f]) if isinstance(args[1], int): # number of anchors args[1] = [list(range(args[1] * 2))] * len(f) diff --git a/models/yolov5s6_decoupled_light.yaml b/models/yolov5s6_decoupled_light.yaml new file mode 100644 index 0000000..39a255c --- /dev/null +++ b/models/yolov5s6_decoupled_light.yaml @@ -0,0 +1,62 @@ +# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license + +# Parameters +nc: 100 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: + - [19, 27, 44, 40, 38, 94] # P3/8 + - [96, 68, 86, 152, 180, 137] # P4/16 + - [140, 301, 303, 264, 238, 542] # P5/32 + - [436, 615, 739, 380, 925, 792] # P6/64 + +# 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, 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, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: [ + [-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, "nearest"]], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, "nearest"]], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, "nearest"]], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + # [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + [[23, 26, 29, 32], 1, DetectDecoupledLite, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] \ No newline at end of file diff --git a/train_my_gra.sh b/train_my_gra.sh index 774e767..e06572c 100644 --- a/train_my_gra.sh +++ b/train_my_gra.sh @@ -5,5 +5,6 @@ python /data6/zhangw/yolov5/train.py \ --epochs 300 \ --data /data6/zhangw/ZW_gra_project/CUB_100/cub100.yaml \ --weights /data6/zhangw/yolov5/yolov5s6.pt \ - --project /data6/zhangw/ZW_gra_project/yolov5_result \ - --name G0_cub100_yolov5s6_raw_e300 \ No newline at end of file + --project /data6/zhangw/ZW_gra_project/yolov5_result_pretrained \ + --name G8_decoupled_head_light_e300 \ + --cfg /data6/zhangw/yolov5/models/yolov5s6_decoupled_light.yaml \ No newline at end of file