一、本文介绍

本文记录的是利用 TFAM 模块改进 YOLOv12 的多模态融合部分 。

TFAM 模块（Temporal Fusion Attention Module，时序融合注意力模块） 通过在特征提取网络的深层 引入通道 - 空间双分支注意力机制 ，基于时序信息动态生成双模态特征融合权重。该模块 可自适应捕捉跨模态的重要特征，抑制双模态特征中的噪声干扰与无效信息，实现高层语义与低层空间特征的时序关联建模与互补融合 ，为变化检测提供跨模态的精准特征表示。

---

---

二、TFAM模块介绍

Exchanging Dual-Encoder–Decoder: A New Strategy for Change Detection With Semantic Guidance and Spatial Localization

2.1 设计出发点

当前基于深度学习的变化检测模型在进行双时相特征融合时，主要存在以下问题：

• 简单融合方法 ：直接进行元素加减或拼接，易受噪声干扰，难以实现有效融合。
• 卷积增强方法 ：虽通过多尺度卷积减少噪声，但忽略了双时相特征间的时序信息。
• 注意力增强方法 ：通常在通道维度拼接后使用注意力机制，但同样未充分利用时序信息。

为解决上述问题，TFAM（Temporal Fusion Attention Module）模块被提出，其核心目标是 利用时序信息实现双时相特征的有效融合 ，通过对比双时相特征在时间维度上的重要性，提升特征融合的准确性和鲁棒性。

2.2 结构原理

TFAM模块通过 通道注意力 和 空间注意力 两个分支，分别在通道和空间维度上计算双时相特征的权重，进而实现特征融合。其具体结构和工作流程如下：

2.2.1 通道注意力分支

• 特征聚合 ：对输入的双时相特征 $T_1$ 和 $T_2$ ，分别进行全局平均池化（Avgpool）和全局最大池化（Maxpool），聚合空间信息，得到 $S_c=\text{Concat}(Avg(T_1),\text{Max}(T_1),\text{Avg}(T_2),\text{Max}(T_2))$ 。
• 权重计算 ：通过两个1D卷积层（类似ECA模块）生成双时相的通道权重 $W_{c1}$ 和 $W_{c2}$ ，再经Softmax归一化，使权重和为1，公式为：
  $$W_{c1}^{\prime },W_{c2}^{\prime }=\frac{e^{W_{c1}}}{e^{W_{c1}}+e^{W_{c2}}},\frac{e^{W_{c2}}}{e^{W_{c1}}+e^{W_{c2}}}$$
  该过程通过对比通道权重，确定双时相特征在通道维度上的重要部分。

2.2.2 空间注意力分支

• 权重计算 ：采用与通道注意力类似的方法，对双时相特征在空间维度上进行池化和卷积操作，生成空间权重 $W_{s1}^{\prime }$ 和 $W_{s2}^{\prime }$ ，用于衡量空间位置的重要性。

2.2.3 特征融合

• 将通道权重和空间权重相加，得到双时相特征的综合权重，再与原始特征相乘并求和，公式为：
  $$\text{Output}=(W_{c1}^{\prime }+W_{s1}^{\prime })\cdot T_1+(W_{c2}^{\prime }+W_{s2}^{\prime })\cdot T_2$$
  最终输出融合后的特征，通过权重分配保留有用信息，丢弃冗余部分。

2.3 优势

1. 有效利用时序信息: 通过对比双时相特征在通道和空间维度的权重，TFAM能够 捕捉跨时相的重要特征 ，避免了传统方法中忽略时序关联的缺陷。

2. 增强特征融合的准确性

   • 通道与空间联合优化 ：同时在两个维度上进行注意力机制计算，使模型能够从“哪里变”（空间）和“什么类型变”（通道语义）两个层面精准定位变化区域。
   • 动态权重分配 ：Softmax归一化确保双时相特征的权重和为1，避免了简单相加或拼接可能引入的噪声放大问题，提升了融合特征的纯净度。

3. 轻量化与高效性

   • 模块通过1D卷积和池化操作实现，计算复杂度低，且可嵌入到主流网络架构中（如EDED backbone），适用于实时或资源受限的场景。
   • 实验表明，引入TFAM后，模型在LEVIR-CD数据集上的F1分数提升了约0.3%，验证了其有效性。

总结

TFAM模块通过 时序信息驱动的注意力机制 ，解决了双时相特征融合中时序关联不足的问题，实现了更精准的变化区域定位和特征表示。其结构轻量、泛化性强，为变化检测任务提供了一种高效的特征融合解决方案。

论文： https://ieeexplore.ieee.org/document/10296953

三、TFAM的实现代码

TFAM 的实现代码如下：

import torch
import torch.nn as nn
import math

def kernel_size(in_channel):
    """Compute kernel size for one dimension convolution in eca-net"""
    k = int((math.log2(in_channel) + 1) // 2)  # parameters from ECA-net
    if k % 2 == 0:
        return k + 1
    else:
        return k

class TFAM(nn.Module):
    """Fuse two feature into one feature."""

    def __init__(self, in_channel):
        super().__init__()

        self.avg_pool = nn.AdaptiveAvgPool2d(1)
        self.max_pool = nn.AdaptiveMaxPool2d(1)
        self.k = kernel_size(in_channel)
        self.channel_conv1 = nn.Conv1d(4, 1, kernel_size=self.k, padding=self.k // 2)
        self.channel_conv2 = nn.Conv1d(4, 1, kernel_size=self.k, padding=self.k // 2)
        self.spatial_conv1 = nn.Conv2d(4, 1, kernel_size=7, padding=3)
        self.spatial_conv2 = nn.Conv2d(4, 1, kernel_size=7, padding=3)
        self.softmax = nn.Softmax(0)

    def forward(self, x, log=None, module_name=None, img_name=None):
        t1 = x[0]  # 拆分输入元组为 t1
        t2 = x[1]  # 拆分输入元组为 t2
        # channel part
        t1_channel_avg_pool = self.avg_pool(t1)  # b,c,1,1
        t1_channel_max_pool = self.max_pool(t1)  # b,c,1,1
        t2_channel_avg_pool = self.avg_pool(t2)  # b,c,1,1
        t2_channel_max_pool = self.max_pool(t2)  # b,c,1,1

        channel_pool = torch.cat([t1_channel_avg_pool, t1_channel_max_pool,
                                  t2_channel_avg_pool, t2_channel_max_pool],
                                 dim=2).squeeze(-1).transpose(1, 2)  # b,4,c
        t1_channel_attention = self.channel_conv1(channel_pool)  # b,1,c
        t2_channel_attention = self.channel_conv2(channel_pool)  # b,1,c
        channel_stack = torch.stack([t1_channel_attention, t2_channel_attention],
                                    dim=0)  # 2,b,1,c
        channel_stack = self.softmax(channel_stack).transpose(-1, -2).unsqueeze(-1)  # 2,b,c,1,1

        # spatial part
        t1_spatial_avg_pool = torch.mean(t1, dim=1, keepdim=True)  # b,1,h,w
        t1_spatial_max_pool = torch.max(t1, dim=1, keepdim=True)[0]  # b,1,h,w
        t2_spatial_avg_pool = torch.mean(t2, dim=1, keepdim=True)  # b,1,h,w
        t2_spatial_max_pool = torch.max(t2, dim=1, keepdim=True)[0]  # b,1,h,w
        spatial_pool = torch.cat([t1_spatial_avg_pool, t1_spatial_max_pool,
                                  t2_spatial_avg_pool, t2_spatial_max_pool], dim=1)  # b,4,h,w
        t1_spatial_attention = self.spatial_conv1(spatial_pool)  # b,1,h,w
        t2_spatial_attention = self.spatial_conv2(spatial_pool)  # b,1,h,w
        spatial_stack = torch.stack([t1_spatial_attention, t2_spatial_attention], dim=0)  # 2,b,1,h,w
        spatial_stack = self.softmax(spatial_stack)  # 2,b,1,h,w

        # fusion part, add 1 means residual add
        stack_attention = channel_stack + spatial_stack + 1  # 2,b,c,h,w
        fuse = stack_attention[0] * t1 + stack_attention[1] * t2  # b,c,h,w

        return fuse

四、融合步骤

4.1 修改一

① 在 ultralytics/nn/ 目录下新建 AddModules 文件夹用于存放模块代码

② 在 AddModules 文件夹下新建 TFAM.py ，将 第三节 中的代码粘贴到此处

4.2 修改二

在 AddModules 文件夹下新建 __init__.py （已有则不用新建），在文件内导入模块： from .TFAM import *

4.3 修改三

在 ultralytics/nn/modules/tasks.py 文件中，需要在两处位置添加各模块类名称。

首先：导入模块

其次：在 parse_model函数 中注册 TFAM 模块

        elif m in {TFAM}:
            c2 = ch[f[0]]
            args = [c2]

在 DetectionModel 类下，添加如下代码

try:
   m.stride = torch.tensor([s / x.shape[-2] for x in _forward(torch.zeros(1, ch, s, s))])  # forward on CPU
except RuntimeError:
   try:
       self.model.to(torch.device('cuda'))
       m.stride = torch.tensor([s / x.shape[-2] for x in _forward(
           torch.zeros(1, ch, s, s).to(torch.device('cuda')))])  # forward on CUDA
   except RuntimeError as error:
       raise error

并注释这一行

# m.stride = torch.tensor([s / x.shape[-2] for x in _forward(torch.zeros(1, ch, s, s))])  # forward

---

五、yaml模型文件

5.1 中期融合⭐

📌 此模型的修方法是将原本的中期融合中的Concat融合部分换成TFAM，融合骨干部分的多模态信息。

# YOLOv12 🚀, AGPL-3.0 license
# YOLOv12 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# CFG file for YOLOv12-turbo

# Parameters
ch: 6
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov12n.yaml' will call yolov12.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 497 layers, 2,553,904 parameters, 2,553,888 gradients, 6.2 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 497 layers, 9,127,424 parameters, 9,127,408 gradients, 19.7 GFLOPs
  m: [0.50, 1.00, 512] # summary: 533 layers, 19,670,784 parameters, 19,670,768 gradients, 60.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 895 layers, 26,506,496 parameters, 26,506,480 gradients, 83.3 GFLOPs
  x: [1.00, 1.50, 512] # summary: 895 layers, 59,414,176 parameters, 59,414,160 gradients, 185.9 GFLOPs

# YOLO12 backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, IN, []]  # 0
  - [-1, 1, Multiin, [1]]  # 1
  - [-2, 1, Multiin, [2]]  # 2

  - [1, 1, Conv,  [64, 3, 2]] # 3-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 4-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 6-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 8-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 10-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 11

  - [2, 1, Conv,  [64, 3, 2]] # 12-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 13-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 15-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 17-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 19-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 20

  - [[7, 16], 1, TFAM, [512]]  # 21 cat backbone P3
  - [[9, 18], 1, TFAM, [512]]  # 22 cat backbone P4
  - [[11, 20], 1, TFAM, [1024]]  # 23 cat backbone P5

# YOLO12 head
head:
  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 22], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, A2C2f, [512, False, -1]] # 26

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 21], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, A2C2f, [256, False, -1]] # 29

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 26], 1, Concat, [1]] # cat head P4
  - [-1, 2, A2C2f, [512, False, -1]] # 32

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 23], 1, Concat, [1]] # cat head P5
  - [-1, 2, C3k2, [1024, True]] # 35 (P5/32-large)

  - [[29, 32, 35], 1, Detect, [nc]] # Detect(P3, P4, P5)

5.2 中-后期融合⭐

📌 此模型的修方法是将原本的中-后期融合中的Concat融合部分换成TFAM，融合FPN部分的多模态信息。

# YOLOv12 🚀, AGPL-3.0 license
# YOLOv12 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# CFG file for YOLOv12-turbo

# Parameters
ch: 6
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov12n.yaml' will call yolov12.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 497 layers, 2,553,904 parameters, 2,553,888 gradients, 6.2 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 497 layers, 9,127,424 parameters, 9,127,408 gradients, 19.7 GFLOPs
  m: [0.50, 1.00, 512] # summary: 533 layers, 19,670,784 parameters, 19,670,768 gradients, 60.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 895 layers, 26,506,496 parameters, 26,506,480 gradients, 83.3 GFLOPs
  x: [1.00, 1.50, 512] # summary: 895 layers, 59,414,176 parameters, 59,414,160 gradients, 185.9 GFLOPs

# YOLO12 backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, IN, []]  # 0
  - [-1, 1, Multiin, [1]]  # 1
  - [-2, 1, Multiin, [2]]  # 2

  - [1, 1, Conv,  [64, 3, 2]] # 3-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 4-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 6-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 8-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 10-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 11

  - [2, 1, Conv,  [64, 3, 2]] # 12-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 13-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 15-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 17-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 19-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 20

# YOLO12 head
head:
  - [11, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 9], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, A2C2f, [512, False, -1]] # 23

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 7], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, A2C2f, [256, False, -1]] # 26

  - [20, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 18], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, A2C2f, [512, False, -1]] # 29

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 16], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, A2C2f, [256, False, -1]] # 32

  - [[11, 20], 1, TFAM, [1024]]  # cat head P5  33
  - [[23, 29], 1, TFAM, [512]]  # cat head P5  34
  - [[26, 32], 1, TFAM, [256]]  # cat head P5  35

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 34], 1, Concat, [1]] # cat head P4
  - [-1, 2, A2C2f, [512, False, -1]] # 38

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 33], 1, Concat, [1]] # cat head P5
  - [-1, 2, C3k2, [1024, True]] # 41 (P5/32-large)

  - [[35, 38, 41], 1, Detect, [nc]] # Detect(P3, P4, P5)

5.3 后期融合⭐

📌 此模型的修方法是将原本的后期融合中的Concat融合部分换成TFAM，融合颈部部分的多模态信息。

# YOLOv12 🚀, AGPL-3.0 license
# YOLOv12 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# CFG file for YOLOv12-turbo

# Parameters
ch: 6
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov12n.yaml' will call yolov12.yaml with scale 'n'
  # [depth, width, max_channels]
  n: [0.50, 0.25, 1024] # summary: 497 layers, 2,553,904 parameters, 2,553,888 gradients, 6.2 GFLOPs
  s: [0.50, 0.50, 1024] # summary: 497 layers, 9,127,424 parameters, 9,127,408 gradients, 19.7 GFLOPs
  m: [0.50, 1.00, 512] # summary: 533 layers, 19,670,784 parameters, 19,670,768 gradients, 60.4 GFLOPs
  l: [1.00, 1.00, 512] # summary: 895 layers, 26,506,496 parameters, 26,506,480 gradients, 83.3 GFLOPs
  x: [1.00, 1.50, 512] # summary: 895 layers, 59,414,176 parameters, 59,414,160 gradients, 185.9 GFLOPs

# YOLO12 backbone
backbone:
  # [from, repeats, module, args]
  - [-1, 1, IN, []]  # 0
  - [-1, 1, Multiin, [1]]  # 1
  - [-2, 1, Multiin, [2]]  # 2

  - [1, 1, Conv,  [64, 3, 2]] # 3-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 4-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 6-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 8-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 10-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 11

  - [2, 1, Conv,  [64, 3, 2]] # 12-P1/2
  - [-1, 1, Conv,  [128, 3, 2, 1, 2]] # 13-P2/4
  - [-1, 2, C3k2,  [256, False, 0.25]]
  - [-1, 1, Conv,  [256, 3, 2, 1, 4]] # 15-P3/8
  - [-1, 2, C3k2,  [512, False, 0.25]]
  - [-1, 1, Conv,  [512, 3, 2]] # 17-P4/16
  - [-1, 4, A2C2f, [512, True, 4]]
  - [-1, 1, Conv,  [1024, 3, 2]] # 19-P5/32
  - [-1, 4, A2C2f, [1024, True, 1]] # 20

# YOLO12 head
head:
  - [11, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 9], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, A2C2f, [512, False, -1]] # 23

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 7], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, A2C2f, [256, False, -1]] # 26

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 23], 1, Concat, [1]] # cat head P4
  - [-1, 2, A2C2f, [512, False, -1]] # 29

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 11], 1, Concat, [1]] # cat head P5
  - [-1, 2, C3k2, [1024, True]] # 32 (P5/32-large)

  - [20, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 18], 1, Concat, [1]] # cat backbone P4
  - [-1, 2, A2C2f, [512, False, -1]] # 35

  - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  - [[-1, 16], 1, Concat, [1]] # cat backbone P3
  - [-1, 2, A2C2f, [256, False, -1]] # 38

  - [-1, 1, Conv, [256, 3, 2]]
  - [[-1, 35], 1, Concat, [1]] # cat head P4
  - [-1, 2, A2C2f, [512, False, -1]] # 41

  - [-1, 1, Conv, [512, 3, 2]]
  - [[-1, 20], 1, Concat, [1]] # cat head P5
  - [-1, 2, C3k2, [1024, True]] # 44 (P5/32-large)

  - [[26, 38], 1, TFAM, [256]]  # cat head P5  45
  - [[29, 41], 1, TFAM, [512]]  # cat head P5  46
  - [[32, 44], 1, TFAM, [1024]]  # cat head P5  47

  - [[45, 46, 47], 1, Detect, [nc]] # Detect(P3, P4, P5)

---

六、成功运行结果

打印网络模型可以看到不同的融合层已经加入到模型中，并可以进行训练了。

YOLOv12-mid-TFAM ：

YOLOv12-mid-TFAM summary: 734 layers, 3,897,903 parameters, 3,897,887 gradients, 8.9 GFLOPs

                   from  n    params  module                                       arguments
  0                  -1  1         0  ultralytics.nn.AddModules.multimodal.IN      []
  1                  -1  1         0  ultralytics.nn.AddModules.multimodal.Multiin [1]
  2                  -2  1         0  ultralytics.nn.AddModules.multimodal.Multiin [2]
  3                   1  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
  4                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
  5                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
  6                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
  7                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
  8                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
  9                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 10                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 11                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 12                   2  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
 13                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
 14                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
 15                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
 16                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
 17                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 18                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 19                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 20                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 21             [7, 16]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [128]
 22             [9, 18]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [128]
 23            [11, 20]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [256]
 24                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 25            [-1, 22]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 26                  -1  1     86912  ultralytics.nn.AddModules.A2C2f.A2C2f        [384, 128, 1, False, -1]
 27                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 28            [-1, 21]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 29                  -1  1     24000  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 64, 1, False, -1]
 30                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]
 31            [-1, 26]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 32                  -1  1     74624  ultralytics.nn.AddModules.A2C2f.A2C2f        [192, 128, 1, False, -1]
 33                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 34            [-1, 23]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 35                  -1  1    378880  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True]
 36        [29, 32, 35]  1    430867  ultralytics.nn.modules.head.Detect           [1, [64, 128, 256]]
YOLOv12-mid-TFAM summary: 734 layers, 3,897,903 parameters, 3,897,887 gradients, 8.9 GFLOPs

YOLOv12-mid-to-late-TFAM ：

YOLOv12-mid-to-late-TFAM summary: 804 layers, 4,008,799 parameters, 4,008,783 gradients, 9.5 GFLOPs

                   from  n    params  module                                       arguments
  0                  -1  1         0  ultralytics.nn.AddModules.multimodal.IN      []
  1                  -1  1         0  ultralytics.nn.AddModules.multimodal.Multiin [1]
  2                  -2  1         0  ultralytics.nn.AddModules.multimodal.Multiin [2]
  3                   1  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
  4                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
  5                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
  6                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
  7                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
  8                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
  9                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 10                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 11                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 12                   2  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
 13                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
 14                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
 15                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
 16                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
 17                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 18                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 19                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 20                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 21                  11  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 22             [-1, 9]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 23                  -1  1     86912  ultralytics.nn.AddModules.A2C2f.A2C2f        [384, 128, 1, False, -1]
 24                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 25             [-1, 7]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 26                  -1  1     24000  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 64, 1, False, -1]
 27                  20  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 28            [-1, 18]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 29                  -1  1     86912  ultralytics.nn.AddModules.A2C2f.A2C2f        [384, 128, 1, False, -1]
 30                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 31            [-1, 16]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 32                  -1  1     24000  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 64, 1, False, -1]
 33            [11, 20]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [256]
 34            [23, 29]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [128]
 35            [26, 32]  1       420  ultralytics.nn.AddModules.TFAM.TFAM          [64]
 36                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]
 37            [-1, 34]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 38                  -1  1     74624  ultralytics.nn.AddModules.A2C2f.A2C2f        [192, 128, 1, False, -1]
 39                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 40            [-1, 33]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 41                  -1  1    378880  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True]
 42        [35, 38, 41]  1    430867  ultralytics.nn.modules.head.Detect           [1, [64, 128, 256]]
YOLOv12-mid-to-late-TFAM summary: 804 layers, 4,008,799 parameters, 4,008,783 gradients, 9.5 GFLOPs

YOLOv12-late-TFAM ：

YOLOv12-late-TFAM summary: 878 layers, 4,647,007 parameters, 4,646,991 gradients, 10.3 GFLOPs

                   from  n    params  module                                       arguments
  0                  -1  1         0  ultralytics.nn.AddModules.multimodal.IN      []
  1                  -1  1         0  ultralytics.nn.AddModules.multimodal.Multiin [1]
  2                  -2  1         0  ultralytics.nn.AddModules.multimodal.Multiin [2]
  3                   1  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
  4                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
  5                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
  6                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
  7                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
  8                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
  9                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 10                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 11                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 12                   2  1       464  ultralytics.nn.modules.conv.Conv             [3, 16, 3, 2]
 13                  -1  1      2368  ultralytics.nn.modules.conv.Conv             [16, 32, 3, 2, 1, 2]
 14                  -1  1      6640  ultralytics.nn.modules.block.C3k2            [32, 64, 1, False, 0.25]
 15                  -1  1      9344  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2, 1, 4]
 16                  -1  1     26080  ultralytics.nn.modules.block.C3k2            [64, 128, 1, False, 0.25]
 17                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 18                  -1  2    180864  ultralytics.nn.AddModules.A2C2f.A2C2f        [128, 128, 2, True, 4]
 19                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]
 20                  -1  2    689408  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 256, 2, True, 1]
 21                  11  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 22             [-1, 9]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 23                  -1  1     86912  ultralytics.nn.AddModules.A2C2f.A2C2f        [384, 128, 1, False, -1]
 24                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 25             [-1, 7]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 26                  -1  1     24000  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 64, 1, False, -1]
 27                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]
 28            [-1, 23]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 29                  -1  1     74624  ultralytics.nn.AddModules.A2C2f.A2C2f        [192, 128, 1, False, -1]
 30                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 31            [-1, 11]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 32                  -1  1    378880  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True]
 33                  20  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 34            [-1, 18]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 35                  -1  1     86912  ultralytics.nn.AddModules.A2C2f.A2C2f        [384, 128, 1, False, -1]
 36                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']
 37            [-1, 16]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 38                  -1  1     24000  ultralytics.nn.AddModules.A2C2f.A2C2f        [256, 64, 1, False, -1]
 39                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]
 40            [-1, 35]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 41                  -1  1     74624  ultralytics.nn.AddModules.A2C2f.A2C2f        [192, 128, 1, False, -1]
 42                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]
 43            [-1, 20]  1         0  ultralytics.nn.modules.conv.Concat           [1]
 44                  -1  1    378880  ultralytics.nn.modules.block.C3k2            [384, 256, 1, True]
 45            [26, 38]  1       420  ultralytics.nn.AddModules.TFAM.TFAM          [64]
 46            [29, 41]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [128]
 47            [32, 44]  1       436  ultralytics.nn.AddModules.TFAM.TFAM          [256]
 48        [45, 46, 47]  1    430867  ultralytics.nn.modules.head.Detect           [1, [64, 128, 256]]
YOLOv12-late-TFAM summary: 878 layers, 4,647,007 parameters, 4,646,991 gradients, 10.3 GFLOPs