update

configs/attention/yolov5s_acmix.yaml

@@ -24,7 +24,7 @@ backbone:
    [-1, 9, C3, [512]],                     
    [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
    [-1, 3, C3, [1024]], 
-   [-1, 1, ACmix, [1024, 1024]], #9
+   [-1, 1, ACmix, [1024]], #9
    [-1, 1, SPPF, [1024,5]], #10
   ]
 

configs/backbone/yolov5_mobileone_block.yaml

@@ -0,0 +1,49 @@
+# YOLOv5改进 🚀 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 backbone by yoloair
+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, 3, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],  # 5-P4/16
+   [-1, 3, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32
+   [-1, 6, MobileOneBlock, [1024, 4, 1, False]],
+   [-1, 1, SPPF, [1024, 5]],  # 9
+  ]
+
+# YOLOv5 head
+head:
+  [[-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]],  # 13
+
+   [-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]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 20 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]
+

configs/yolov5_exp/yolov5_repvgg_v6backbone.yaml

@@ -0,0 +1,48 @@
+# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
+
+# Parameters
+nc: 80  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+anchors:
+  - [10,13, 16,30, 33,23]  # P3/8
+  - [30,61, 62,45, 59,119]  # P4/16
+  - [116,90, 156,198, 373,326]  # P5/32
+
+# YOLOv5 v6.0 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, RepVGGBlockv6,[64,3,2]],  # 0-P1/2
+   [-1, 1, RepVGGBlockv6,[128,3,2]],  # 1-P2/4
+   [-1, 3, RepBlock,[128]],
+   [-1, 1, RepVGGBlockv6,[256,3,2]],  # 3-P3/8
+   [-1, 6, RepBlock,[256]],
+   [-1, 1, RepVGGBlockv6,[512,3,2]],  # 5-P4/16
+   [-1, 9, RepBlock,[512]],
+   [-1, 1, RepVGGBlockv6,[1024,3,2]],  # 7-P5/32
+   [-1, 3, RepBlock,[1024]],
+   [-1, 1, SimSPPF,[1024]],  # 9
+  ]
+
+# YOLOv5 v6.0 head
+head:
+  [[-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]],  # 13
+
+   [-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]],  # 17 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 14], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 10], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 23 (P5/32-large)
+
+   [[17, 20, 23], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/Models/yolov4.py

@@ -56,7 +56,7 @@ def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, k
         super(Conv, self).__init__()
         self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
         self.bn = nn.BatchNorm2d(c2)
-        self.act = Mish() if act else nn.Identity()
+        self.act = nn.Mish() if act else nn.Identity()
 
     def forward(self, x):
         return self.act(self.bn(self.conv(x)))
@@ -88,7 +88,7 @@ def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, nu
         self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
         self.cv4 = Conv(2 * c_, c2, 1, 1)
         self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
-        self.act = Mish()
+        self.act = nn.Mish()
         self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
 
     def forward(self, x):
@@ -107,7 +107,7 @@ def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, nu
         #self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
         self.cv4 = Conv(2 * c_, c2, 1, 1)
         self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
-        self.act = Mish()
+        self.act = nn.Mish()
         self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
 
     def forward(self, x):
@@ -126,7 +126,7 @@ def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, nu
         self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
         #self.cv4 = Conv(2 * c_, c2, 1, 1)
         self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
-        self.act = Mish()
+        self.act = nn.Mish()
         self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
 
     def forward(self, x):
@@ -144,7 +144,7 @@ def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5):  # ch_in, ch_out, n
         self.cv2 = nn.Conv2d(c_, c_, 1, 1, bias=False)
         self.cv3 = Conv(2 * c_, c2, 1, 1)
         self.bn = nn.BatchNorm2d(2 * c_) 
-        self.act = Mish()
+        self.act = nn.Mish()
         self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
 
     def forward(self, x):
@@ -197,7 +197,7 @@ def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5, k=(5, 9, 13)):
         self.cv5 = Conv(4 * c_, c_, 1, 1)
         self.cv6 = Conv(c_, c_, 3, 1)
         self.bn = nn.BatchNorm2d(2 * c_) 
-        self.act = Mish()
+        self.act = nn.Mish()
         self.cv7 = Conv(2 * c_, c2, 1, 1)
 
     def forward(self, x):
@@ -430,8 +430,9 @@ def forward_fuse(self, x):
 
 class SimSPPF(nn.Module):
     '''Simplified SPPF with ReLU activation'''
-    def __init__(self, in_channels, out_channels, kernel_size=5):
+    def __init__(self, in_channels, kernel_size=5):
         super().__init__()
+        out_channels = in_channels
         c_ = in_channels // 2  # hidden channels
         self.cv1 = SimConv(in_channels, c_, 1, 1)
         self.cv2 = SimConv(c_ * 4, out_channels, 1, 1)

models/yolo.py

@@ -347,7 +347,7 @@ def parse_model(d, ch):  # model_dict, input_channels(3)
 
         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, C3HB, C3RFEM, MultiSEAM, SEAM, C3STR]:
+                 BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, C3HB, C3RFEM, MultiSEAM, SEAM, C3STR, MobileOneBlock]:
             c1, c2 = ch[f], args[0]
             if c2 != no:  # if not output
                 c2 = make_divisible(c2 * gw, 8)
@@ -357,9 +357,9 @@ def parse_model(d, ch):  # model_dict, input_channels(3)
                 args.insert(2, n)  # number of repeats
                 n = 1
         # add module research
-        elif m in [CARAFE, SPPCSPC, SPPFCSPC, RepConv, BoT3, CA, CBAM, NAMAttention, GAMAttention, ACmix, Involution, Stem, ResCSPC, ResCSPB, \
+        elif m in [CARAFE, SPPCSPC, SPPFCSPC, RepConv, BoT3, CA, CBAM, NAMAttention, GAMAttention, Involution, Stem, ResCSPC, ResCSPB, \
                    ResXCSPB, ResXCSPC, BottleneckCSPB, BottleneckCSPC,
-                   ASPP, BasicRFB, SPPCSPC_group, SimSPPF, HorBlock, CNeB,C3GC ,C3C2, nn.ConvTranspose2d]:
+                   ASPP, BasicRFB, SPPCSPC_group, HorBlock, CNeB,C3GC ,C3C2, nn.ConvTranspose2d]:
             c1, c2 = ch[f], args[0]
             if c2 != no:  # if not output
                 c2 = make_divisible(c2 * gw, 8)
@@ -374,7 +374,7 @@ def parse_model(d, ch):  # model_dict, input_channels(3)
                     args[6] = make_divisible(args[6] * gw, 8)
         elif m in [CBH, ES_Bottleneck, DWConvblock, RepVGGBlock, LC_Block, Dense, conv_bn_relu_maxpool, \
                    Shuffle_Block, stem, mobilev3_bneck, conv_bn_hswish, MobileNetV3_InvertedResidual, DepthSepConv, \
-                   ShuffleNetV2_Model, Conv_maxpool, CoT3, ConvNextBlock]:
+                   ShuffleNetV2_Model, Conv_maxpool, CoT3, ConvNextBlock, RepBlock]:
             c1, c2 = ch[f], args[0]
             if c2 != no:  # if not output
                 c2 = make_divisible(c2 * gw, 8)
@@ -395,7 +395,7 @@ def parse_model(d, ch):  # model_dict, input_channels(3)
                 n = 1
         elif m in [ReOrg, DWT]:
             c2 = ch[f] * 4
-        elif m in [S2Attention, CrissCrossAttention, SOCA, ShuffleAttention, SEAttention, SimAM, SKAttention]:
+        elif m in [S2Attention, SimSPPF, ACmix, CrissCrossAttention, SOCA, ShuffleAttention, SEAttention, SimAM, SKAttention]:
             c1, c2 = ch[f], args[0]
             if c2 != no:  # if not output
                 c2 = make_divisible(c2 * gw, 8)

train.py

@@ -53,7 +53,7 @@
 from utils.loggers import Loggers
 from utils.loggers.wandb.wandb_utils import check_wandb_resume
 from utils.loss import ComputeLoss, ComputeNWDLoss, ComputeLossOTA_v7, ComputeLossAuxOTA, ComputeLossBinOTA
-# from utils.loss_ps import ComputeLoss_v4
+from utils.loss_ps import ComputeLoss_v4
 from utils.metrics import fitness
 from utils.plots import plot_evolve, plot_labels
 from utils.torch_utils import EarlyStopping, ModelEMA, de_parallel, is_parallel, select_device, torch_distributed_zero_first
@@ -290,6 +290,8 @@ def train(hyp,  # path/to/hyp.yaml or hyp dictionary
         compute_loss = ComputeLoss(model)
     if loss_category is None:
         compute_loss = ComputeLoss(model)  # init loss class
+    elif opt.loss == 'v4':
+        compute_loss = ComputeLoss_v4(model)
     else:
         compute_loss = loss_category(model)# loss class
     if opt.loss == 'nwd':