1. 代码

import torch
import torch.nn as nn

class DyT(nn.Module):
    # input x has the shape of[B,T，C]

    # B:batch size, T:tokens, C:dimension

    def __init__(self, C, init_alpha):
        super(DyT, self).__init__()

        self.alpha = nn.Parameter(torch.ones(1) * init_alpha)
        self.beta = nn.Parameter(torch.ones(C))
        self.gamma = nn.Parameter(torch.zeros(C))
        self.act = nn.Tanh()

    def forward(self, x):

        b, c, h, w = x.shape
        x = x.reshape(b, c, -1).permute(0, 2, 1)

        x = self.act(x * self.alpha)
        x = self.beta * x + self.gamma

        return x.permute(0, 2, 1).reshape(b, c, h, w)
    

if __name__ == '__main__':
    x = torch.randn(1, 100, 512)
    model = DyT(512, 0.1)
    y = model(x)
    print(y.shape)

2. yaml配置文件

2.1 添加三层DyT模块

# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLO11 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect

# Parameters
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolo11n.yaml' will call yolo11.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 319 layers, 2624080 parameters, 2624064 gradients, 6.6 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 319 layers, 9458752 parameters, 9458736 gradients, 21.7 GFLOPs
  m: [0.50, 1.00, 512] # summary: 409 layers, 20114688 parameters, 20114672 gradients, 68.5 GFLOPs
  l: [1.00, 1.00, 512] # summary: 631 layers, 25372160 parameters, 25372144 gradients, 87.6 GFLOPs
  x: [1.00, 1.50, 512] # summary: 631 layers, 56966176 parameters, 56966160 gradients, 196.0 GFLOPs

# YOLO11n backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, Conv, [64, 3, 2]] # 0-P1/2
  - [-1, 1, Conv, [128, 3, 2]] # 1-P2/4
  - [-1, 2, C3k2, [256, False, 0.25]]
  - [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
  - [-1, 2, C3k2, [512, False, 0.25]]
  - [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
  - [-1, 2, C3k2, [512, True]]
  - [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
  - [-1, 2, C3k2, [1024, True]]
  - [-1, 1, SPPF, [1024, 5]] # 9
  - [-1, 2, C2PSA, [1024]] # 10

# YOLO11n head
head:
  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 6], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, C3k2, [512, False]] # 13

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 4], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, C3k2, [256, False]] # 16 (P3/8-small)
  - [-1, 1, DyT, [64, 0.8]]

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 13], 1, Concat, [1]] # cat head P4
  - [-1, 2, C3k2, [512, False]] # 19 (P4/16-medium)
  - [-1, 1, DyT, [128, 0.8]]

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 10], 1, Concat, [1]] # cat head P5
  - [-1, 2, C3k2, [1024, True]] # 22 (P5/32-large)
  - [-1, 1, DyT, [256, 0.8]]

  - [[16, 19, 22], 1, Detect, [nc]] # Detect(P3, P4, P5)

2.2 改进DyT-C3k2模块

打开 ultralytics\nn\modules\block.py文件，导入DyT类，修改Bottleneck类

class Bottleneck(nn.Module):
    """Standard bottleneck."""

    def __init__(self, c1, c2, shortcut=True, g=1, k=(3, 3), e=0.5):
        """Initializes a standard bottleneck module with optional shortcut connection and configurable parameters."""
        super().__init__()
        c_ = int(c2 * e)  # hidden channels
        self.cv1 = Conv(c1, c_, k[0], 1)
        self.cv2 = Conv(c_, c2, k[1], 1, g=g)
        self.add = shortcut and c1 == c2
        self.dyt = DyT(c2, 0.85)
        
    def forward(self, x):
        """Applies the YOLO FPN to input data."""
        return x + self.dyt(self.cv2(self.cv1(x))) if self.add else self.dyt(self.cv2(self.cv1(x)))

3. 如何使用

打开 ultralytics\nn\task.py文件，导入DyT类

在parse_model函数进行修改。

4.实验结果

改进前

改进后