一、本文介绍
本文记录的是利用 PSFM 模块改进 YOLOv13 的多模态融合部分 。
PSFM模块（Profound Semantic Fusion Module，深层语义融合模块） 通过在特征提取网络的深层引入 跨模态交叉注意力机制，动态建模红外与可见光特征的全局语义依赖关系 。该模块以可见光语义为引导、红外热信号为补充，通过 双向注意力 计算实现“语义类别”与“热目标位置”的精准对齐，同时 捕捉长距离语义关联 ， 增强融合特征的判别性与场景理解能力 ，为检测头 提供包含全局上下文的高层语义表示 ，从而提升模型在复杂场景下的目标检测准确率与语义推理鲁棒性。
文章目录
二、PSFM模块介绍
Rethinking the necessity of image fusion in high-level vision tasks: A practical infrared and visible image fusion network based on progressive semantic injection and scene fidelity
2.1 设计目标
- 解决高层语义融合的跨模态鸿沟 ：红外图像的热信号与可见光图像的语义类别存在模态差异（如红外“热斑”对应可见光“行人”），需通过语义交互建立跨模态映射。
- 增强全局语义一致性 ：确保融合特征中目标的语义标签（如类别）与可见光图像一致，同时保留红外目标的空间位置信息，避免语义混淆（如将车辆热信号误判为行人）。
- 支持复杂场景的语义推理 ：通过全局上下文建模，捕捉长距离依赖关系（如“行人-道路”的空间关系），提升模型对复杂场景的理解能力。
2.2 结构原理：基于交叉注意力的全局语义交互
2.2.1 核心组件与流程
PSFM模块的架构如图所示，基于 跨模态交叉注意力机制 （Cross-Attention）实现深层特征的语义融合，主要步骤如下：
- 特征投影与维度变换
-
对红外和可见光的深层特征分别进行投影，生成注意力机制所需的
键（Key）
、**值（Value）**矩阵：
K x i = Reshape ( Conv K x ( F ^ x i ) ) , V x i = Reshape ( Conv V x ( F ^ x i ) ) K_{x}^{i} = \text{Reshape}\left(\text{Conv}_{K}^{x}\left(\hat{\mathcal{F}}_{x}^{i}\right)\right), \quad V_{x}^{i} = \text{Reshape}\left(\text{Conv}_{V}^{x}\left(\hat{\mathcal{F}}_{x}^{i}\right)\right) K x i = Reshape ( Conv K x ( F ^ x i ) ) , V x i = Reshape ( Conv V x ( F ^ x i ) )
（其中 x ∈ { ir , vi } x \in \{\text{ir}, \text{vi}\} x ∈ { ir , vi } ， Conv K x \text{Conv}_{K}^{x} Conv K x 和 Conv V x \text{Conv}_{V}^{x} Conv V x 为3×3卷积， Reshape \text{Reshape} Reshape 将特征图展开为 H W × C HW \times C H W × C 的矩阵，便于注意力计算）。
- 跨模态注意力计算
-
以可见光特征为
查询（Query, Q）
，计算其与红外特征的键矩阵
K
ir
i
K_{\text{ir}}^{i}
K
ir
i
的注意力矩阵
A
ir
i
\mathcal{A}_{\text{ir}}^{i}
A
ir
i
：
A ir i = Softmax ( Q i ⋅ ( K ir i ) T ) \mathcal{A}_{\text{ir}}^{i} = \text{Softmax}\left(Q^{i} \cdot (K_{\text{ir}}^{i})^{\text{T}}\right) A ir i = Softmax ( Q i ⋅ ( K ir i ) T )
（ A ir i \mathcal{A}_{\text{ir}}^{i} A ir i 表示可见光特征对红外特征的依赖程度，数值越大表明该区域越需要红外语义信息）。 - 同理，计算红外特征对可见光特征的注意力矩阵 A vi i \mathcal{A}_{\text{vi}}^{i} A vi i ，实现双向语义交互。
- 全局语义特征聚合
-
根据注意力矩阵加权聚合跨模态的
值矩阵（Value）
，生成包含全局上下文的特征：
Attn ir i = A ir i ⋅ V ir i , Attn vi i = A vi i ⋅ V vi i \text{Attn}_{\text{ir}}^{i} = \mathcal{A}_{\text{ir}}^{i} \cdot V_{\text{ir}}^{i}, \quad \text{Attn}_{\text{vi}}^{i} = \mathcal{A}_{\text{vi}}^{i} \cdot V_{\text{vi}}^{i} Attn ir i = A ir i ⋅ V ir i , Attn vi i = A vi i ⋅ V vi i -
将聚合后的特征与原始特征相加，并在通道维度拼接，通过卷积层生成最终的融合特征：
F f u i = Conv ( C ( F vi i + Reshape ( Attn ir i ) , F ir i + Reshape ( Attn vi i ) ) ) \mathcal{F}_{fu}^{i} = \text{Conv}\left(\mathcal{C}\left(\mathcal{F}_{\text{vi}}^{i} + \text{Reshape}(\text{Attn}_{\text{ir}}^{i}), \mathcal{F}_{\text{ir}}^{i} + \text{Reshape}(\text{Attn}_{\text{vi}}^{i})\right)\right) F f u i = Conv ( C ( F vi i + Reshape ( Attn ir i ) , F ir i + Reshape ( Attn vi i ) ) )
（ C \mathcal{C} C 表示通道拼接，通过残差连接保留原始特征的语义信息，避免注意力机制导致的特征失真）。
3. 关键技术特点
- 双向跨模态交互 ：通过可见光与红外特征的双向注意力计算（Q分别来自可见光和红外），实现“以可见光语义为引导对齐红外目标”和“以红外热信号增强可见光语义”的双向优化。
- 全局上下文建模 ：注意力机制可捕捉特征图中任意位置的语义依赖关系（如远处行人与近处车辆的关联），解决传统卷积神经网络对长距离依赖建模能力不足的问题。
- 轻量化设计 ：仅通过少量卷积层和矩阵运算实现，参数增量小于5%，适用于深层特征的高效语义融合。
论文： https://www.sciencedirect.com/science/article/abs/pii/S1566253523001860
源码： https://github.com/Linfeng-Tang/PSFusion
三、PSFM的实现代码
PSFM
的实现代码如下：
import math
import torch.nn as nn
import torch
def autopad(k, p=None, d=1): # kernel, padding, dilation
"""Pad to 'same' shape outputs."""
if d > 1:
k = d * (k - 1) + 1 if isinstance(k, int) else [d * (x - 1) + 1 for x in k] # actual kernel-size
if p is None:
p = k // 2 if isinstance(k, int) else [x // 2 for x in k] # auto-pad
return p
class Conv(nn.Module):
"""Standard convolution with args(ch_in, ch_out, kernel, stride, padding, groups, dilation, activation)."""
default_act = nn.SiLU() # default activation
def __init__(self, c1, c2, k=1, s=1, p=None, g=1, d=1, act=True):
"""Initialize Conv layer with given arguments including activation."""
super().__init__()
self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p, d), groups=g, dilation=d, bias=False)
self.bn = nn.BatchNorm2d(c2)
self.act = self.default_act if act is True else act if isinstance(act, nn.Module) else nn.Identity()
def forward(self, x):
"""Apply convolution, batch normalization and activation to input tensor."""
return self.act(self.bn(self.conv(x)))
def forward_fuse(self, x):
"""Perform transposed convolution of 2D data."""
return self.act(self.conv(x))
class DWConv(Conv):
"""Depth-wise convolution."""
def __init__(self, c1, c2, k=1, s=1, d=1, act=True): # ch_in, ch_out, kernel, stride, dilation, activation
"""Initialize Depth-wise convolution with given parameters."""
super().__init__(c1, c2, k, s, g=math.gcd(c1, c2), d=d, act=act)
class DSConv(nn.Module):
"""Depthwise Separable Convolution"""
def __init__(self, c1, c2, k=1, s=1, d=1, act=True) -> None:
super().__init__()
self.dwconv = DWConv(c1, c1, 3)
self.pwconv = Conv(c1, c2, 1)
def forward(self, x):
return self.pwconv(self.dwconv(x))
class GEFM(nn.Module):
def __init__(self, in_C, out_C):
super(GEFM, self).__init__()
self.RGB_K= DSConv(out_C, out_C, 3)
self.RGB_V = DSConv(out_C, out_C, 3)
self.Q = DSConv(in_C, out_C, 3)
self.INF_K= DSConv(out_C, out_C, 3)
self.INF_V = DSConv(out_C, out_C, 3)
self.Second_reduce = DSConv(in_C, out_C, 3)
self.gamma1 = nn.Parameter(torch.zeros(1))
self.gamma2 = nn.Parameter(torch.zeros(1))
self.softmax = nn.Softmax(dim=-1)
def forward(self, x, y):
Q = self.Q(torch.cat([x,y], dim=1))
RGB_K = self.RGB_K(x)
RGB_V = self.RGB_V(x)
m_batchsize, C, height, width = RGB_V.size()
RGB_V = RGB_V.view(m_batchsize, -1, width*height)
RGB_K = RGB_K.view(m_batchsize, -1, width*height).permute(0, 2, 1)
RGB_Q = Q.view(m_batchsize, -1, width*height)
RGB_mask = torch.bmm(RGB_K, RGB_Q)
RGB_mask = self.softmax(RGB_mask)
RGB_refine = torch.bmm(RGB_V, RGB_mask.permute(0, 2, 1))
RGB_refine = RGB_refine.view(m_batchsize, -1, height,width)
RGB_refine = self.gamma1*RGB_refine+y
INF_K = self.INF_K(y)
INF_V = self.INF_V(y)
INF_V = INF_V.view(m_batchsize, -1, width*height)
INF_K = INF_K.view(m_batchsize, -1, width*height).permute(0, 2, 1)
INF_Q = Q.view(m_batchsize, -1, width*height)
INF_mask = torch.bmm(INF_K, INF_Q)
INF_mask = self.softmax(INF_mask)
INF_refine = torch.bmm(INF_V, INF_mask.permute(0, 2, 1))
INF_refine = INF_refine.view(m_batchsize, -1, height,width)
INF_refine = self.gamma2 * INF_refine + x
out = self.Second_reduce(torch.cat([RGB_refine, INF_refine], dim=1))
return out
class DenseLayer(nn.Module):
def __init__(self, in_C, out_C, down_factor=4, k=2):
super(DenseLayer, self).__init__()
self.k = k
self.down_factor = down_factor
mid_C = out_C // self.down_factor
self.down = nn.Conv2d(in_C, mid_C, 1)
self.denseblock = nn.ModuleList()
for i in range(1, self.k + 1):
self.denseblock.append(DSConv(mid_C * i, mid_C, 3))
self.fuse = DSConv(in_C + mid_C, out_C, 3)
def forward(self, in_feat):
down_feats = self.down(in_feat)
out_feats = []
for i in self.denseblock:
feats = i(torch.cat((*out_feats, down_feats), dim=1))
out_feats.append(feats)
feats = torch.cat((in_feat, feats), dim=1)
return self.fuse(feats)
class PSFM(nn.Module):
def __init__(self, Channel):
super(PSFM, self).__init__()
self.RGBobj = DenseLayer(Channel, Channel)
self.Infobj = DenseLayer(Channel, Channel)
self.obj_fuse = GEFM(Channel * 2, Channel)
def forward(self, data):
rgb, depth = data
rgb_sum = self.RGBobj(rgb)
Inf_sum = self.Infobj(depth)
out = self.obj_fuse(rgb_sum,Inf_sum)
return out
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
PSFM.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 .PSFM 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
PSFM
模块
elif m in {PSFM}:
c2 = ch[f[0]]
args = [c2]
最后将
ultralytics/utils/torch_utils.py
中的
get_flops
函数中的
stride
指定为
640
。
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Add融合部分换成PSFM，融合骨干部分的多模态信息。
ch: 6
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov13n.yaml' will call yolov13.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.50, 0.25, 1024] # Nano
s: [0.50, 0.50, 1024] # Small
l: [1.00, 1.00, 512] # Large
x: [1.00, 1.50, 512] # Extra Large
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, DSC3k2, [256, False, 0.25]]
- [-1, 1, Conv, [256, 3, 2, 1, 4]] # 6-P3/8
- [-1, 2, DSC3k2, [512, True]]
- [-1, 1, DSConv, [512, 3, 2]] # 8-P4/16
- [-1, 4, A2C2f, [512, True, 4]]
- [-1, 1, DSConv, [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, DSC3k2, [256, False, 0.25]]
- [-1, 1, Conv, [256, 3, 2, 1, 4]] # 15-P3/8
- [-1, 2, DSC3k2, [512, True]]
- [-1, 1, DSConv, [512, 3, 2]] # 17-P4/16
- [-1, 4, A2C2f, [512, True, 4]]
- [-1, 1, DSConv, [1024, 3, 2]] # 19-P5/32
- [-1, 4, A2C2f, [1024, True, 1]] # 20
- [[7, 16], 1, PSFM, []] # 21 cat backbone P3
- [[9, 18], 1, PSFM, []] # 22 cat backbone P4
- [[11, 20], 1, PSFM, []] # 23 cat backbone P5
head:
- [[21, 22, 23], 2, HyperACE, [512, 8, True, True, 0.5, 1, "both"]]
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [ 24, 1, DownsampleConv, []]
- [[22, 24], 1, FullPAD_Tunnel, []] # 27
- [[21, 25], 1, FullPAD_Tunnel, []] # 28
- [[23, 26], 1, FullPAD_Tunnel, []] # 29
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 27], 1, Concat, [1]] # cat backbone P4
- [-1, 2, DSC3k2, [512, True]] # 32
- [[-1, 24], 1, FullPAD_Tunnel, []] # 33
- [32, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 28], 1, Concat, [1]] # cat backbone P3
- [-1, 2, DSC3k2, [256, True]] # 36
- [25, 1, Conv, [256, 1, 1]]
- [[36, 37], 1, FullPAD_Tunnel, []] # 38
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 33], 1, Concat, [1]] # cat head P4
- [-1, 2, DSC3k2, [512, True]] # 41
- [[-1, 24], 1, FullPAD_Tunnel, []]
- [41, 1, Conv, [512, 3, 2]]
- [[-1, 29], 1, Concat, [1]] # cat head P5
- [-1, 2, DSC3k2, [1024,True]] # 45 (P5/32-large)
- [[-1, 26], 1, FullPAD_Tunnel, []]
- [[38, 42, 46], 1, Detect, [nc]] # Detect(P3, P4, P5)
5.2 后期融合⭐
📌 此模型的修方法是将原本的后期融合中的Add融合部分换成PSFM，融合颈部部分的多模态信息。
ch: 6
nc: 80 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov13n.yaml' will call yolov13.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.50, 0.25, 1024] # Nano
s: [0.50, 0.50, 1024] # Small
l: [1.00, 1.00, 512] # Large
x: [1.00, 1.50, 512] # Extra Large
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, DSC3k2, [256, False, 0.25]]
- [-1, 1, Conv, [256, 3, 2, 1, 4]] # 6-P3/8
- [-1, 2, DSC3k2, [512, True]]
- [-1, 1, DSConv, [512, 3, 2]] # 8-P4/16
- [-1, 4, A2C2f, [512, True, 4]]
- [-1, 1, DSConv, [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, DSC3k2, [256, False, 0.25]]
- [-1, 1, Conv, [256, 3, 2, 1, 4]] # 15-P3/8
- [-1, 2, DSC3k2, [512, True]]
- [-1, 1, DSConv, [512, 3, 2]] # 17-P4/16
- [-1, 4, A2C2f, [512, True, 4]]
- [-1, 1, DSConv, [1024, 3, 2]] # 19-P5/32
- [-1, 4, A2C2f, [1024, True, 1]] # 20
head:
- [[7, 9, 11], 2, HyperACE, [512, 8, True, True, 0.5, 1, "both"]]
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [ 21, 1, DownsampleConv, []]
- [[9, 21], 1, FullPAD_Tunnel, []] # 24
- [[7, 22], 1, FullPAD_Tunnel, []] # 25
- [[11, 23], 1, FullPAD_Tunnel, []] # 26
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 24], 1, Concat, [1]] # cat backbone P4
- [-1, 2, DSC3k2, [512, True]] # 29
- [[-1, 21], 1, FullPAD_Tunnel, []] # 30
- [29, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 25], 1, Concat, [1]] # cat backbone P3
- [-1, 2, DSC3k2, [256, True]] # 33
- [22, 1, Conv, [256, 1, 1]]
- [[33, 34], 1, FullPAD_Tunnel, []] # 35
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 30], 1, Concat, [1]] # cat head P4
- [-1, 2, DSC3k2, [512, True]] # 38
- [[-1, 21], 1, FullPAD_Tunnel, []]
- [38, 1, Conv, [512, 3, 2]]
- [[-1, 26], 1, Concat, [1]] # cat head P5
- [-1, 2, DSC3k2, [1024,True]] # 42 (P5/32-large)
- [[-1, 23], 1, FullPAD_Tunnel, []]
- [[16, 18, 20], 2, HyperACE, [512, 8, True, True, 0.5, 1, "both"]]
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [ 44, 1, DownsampleConv, []]
- [[18, 44], 1, FullPAD_Tunnel, []] # 47
- [[16, 45], 1, FullPAD_Tunnel, []] # 48
- [[20, 46], 1, FullPAD_Tunnel, []] # 49
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 47], 1, Concat, [1]] # cat backbone P4
- [-1, 2, DSC3k2, [512, True]] # 52
- [[-1, 44], 1, FullPAD_Tunnel, []] # 53
- [52, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 48], 1, Concat, [1]] # cat backbone P3
- [-1, 2, DSC3k2, [256, True]] # 56
- [45, 1, Conv, [256, 1, 1]]
- [[56, 57], 1, FullPAD_Tunnel, []] # 58
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 53], 1, Concat, [1]] # cat head P4
- [-1, 2, DSC3k2, [512, True]] # 61
- [[-1, 44], 1, FullPAD_Tunnel, []]
- [61, 1, Conv, [512, 3, 2]]
- [[-1, 49], 1, Concat, [1]] # cat head P5
- [-1, 2, DSC3k2, [1024,True]] # 65 (P5/32-large)
- [[-1, 46], 1, FullPAD_Tunnel, []]
- [[35, 58], 1, PSFM, []] # 67 cat backbone P3
- [[39, 62], 1, PSFM, []] # 68 cat backbone P4
- [[43, 66], 1, PSFM, []] # 69 cat backbone P5
- [[67, 68, 69], 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。
YOLOv13-mid-PSFM ：
YOLOv13-mid-PSFM summary: 1,197 layers, 4,743,768 parameters, 4,743,752 gradients, 14.4 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 5792 ultralytics.nn.modules.block.DSC3k2 [32, 64, 1, False, 0.25]
6 -1 1 9344 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2, 1, 4]
7 -1 1 74368 ultralytics.nn.modules.block.DSC3k2 [64, 128, 1, True]
8 -1 1 17792 ultralytics.nn.modules.conv.DSConv [128, 128, 3, 2]
9 -1 2 180864 ultralytics.nn.AddModules.A2C2f.A2C2f [128, 128, 2, True, 4]
10 -1 1 34432 ultralytics.nn.modules.conv.DSConv [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 5792 ultralytics.nn.modules.block.DSC3k2 [32, 64, 1, False, 0.25]
15 -1 1 9344 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2, 1, 4]
16 -1 1 74368 ultralytics.nn.modules.block.DSC3k2 [64, 128, 1, True]
17 -1 1 17792 ultralytics.nn.modules.conv.DSConv [128, 128, 3, 2]
18 -1 2 180864 ultralytics.nn.AddModules.A2C2f.A2C2f [128, 128, 2, True, 4]
19 -1 1 34432 ultralytics.nn.modules.conv.DSConv [128, 256, 3, 2]
20 -1 2 689408 ultralytics.nn.AddModules.A2C2f.A2C2f [256, 256, 2, True, 1]
21 [7, 16] 1 205634 ultralytics.nn.AddModules.PSFM.PSFM [128]
22 [9, 18] 1 205634 ultralytics.nn.AddModules.PSFM.PSFM [128]
23 [11, 20] 1 784002 ultralytics.nn.AddModules.PSFM.PSFM [256]
24 [21, 22, 23] 1 273536 ultralytics.nn.modules.block.HyperACE [128, 128, 1, 4, True, True, 0.5, 1, 'both']
25 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
26 24 1 33280 ultralytics.nn.modules.block.DownsampleConv [128]
27 [22, 24] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
28 [21, 25] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
29 [23, 26] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
30 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
31 [-1, 27] 1 0 ultralytics.nn.modules.conv.Concat [1]
32 -1 1 115328 ultralytics.nn.modules.block.DSC3k2 [384, 128, 1, True]
33 [-1, 24] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
34 32 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
35 [-1, 28] 1 0 ultralytics.nn.modules.conv.Concat [1]
36 -1 1 35136 ultralytics.nn.modules.block.DSC3k2 [256, 64, 1, True]
37 25 1 8320 ultralytics.nn.modules.conv.Conv [128, 64, 1, 1]
38 [36, 37] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
39 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
40 [-1, 33] 1 0 ultralytics.nn.modules.conv.Concat [1]
41 -1 1 90752 ultralytics.nn.modules.block.DSC3k2 [192, 128, 1, True]
42 [-1, 24] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
43 41 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
44 [-1, 29] 1 0 ultralytics.nn.modules.conv.Concat [1]
45 -1 1 345344 ultralytics.nn.modules.block.DSC3k2 [384, 256, 1, True]
46 [-1, 26] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
47 [38, 42, 46] 1 432427 ultralytics.nn.modules.head.Detect [9, [64, 128, 256]]
YOLOv13-mid-PSFM summary: 1,197 layers, 4,743,768 parameters, 4,743,752 gradients, 14.4 GFLOPs
YOLOv13-late-PSFM ：
YOLOv13-late-PSFM summary: 1,480 layers, 5,680,767 parameters, 5,680,751 gradients, 14.9 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 5792 ultralytics.nn.modules.block.DSC3k2 [32, 64, 1, False, 0.25]
6 -1 1 9344 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2, 1, 4]
7 -1 1 74368 ultralytics.nn.modules.block.DSC3k2 [64, 128, 1, True]
8 -1 1 17792 ultralytics.nn.modules.conv.DSConv [128, 128, 3, 2]
9 -1 2 180864 ultralytics.nn.AddModules.A2C2f.A2C2f [128, 128, 2, True, 4]
10 -1 1 34432 ultralytics.nn.modules.conv.DSConv [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 5792 ultralytics.nn.modules.block.DSC3k2 [32, 64, 1, False, 0.25]
15 -1 1 9344 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2, 1, 4]
16 -1 1 74368 ultralytics.nn.modules.block.DSC3k2 [64, 128, 1, True]
17 -1 1 17792 ultralytics.nn.modules.conv.DSConv [128, 128, 3, 2]
18 -1 2 180864 ultralytics.nn.AddModules.A2C2f.A2C2f [128, 128, 2, True, 4]
19 -1 1 34432 ultralytics.nn.modules.conv.DSConv [128, 256, 3, 2]
20 -1 2 689408 ultralytics.nn.AddModules.A2C2f.A2C2f [256, 256, 2, True, 1]
21 [7, 9, 11] 1 273536 ultralytics.nn.modules.block.HyperACE [128, 128, 1, 4, True, True, 0.5, 1, 'both']
22 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
23 21 1 33280 ultralytics.nn.modules.block.DownsampleConv [128]
24 [9, 21] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
25 [7, 22] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
26 [11, 23] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
27 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
28 [-1, 24] 1 0 ultralytics.nn.modules.conv.Concat [1]
29 -1 1 115328 ultralytics.nn.modules.block.DSC3k2 [384, 128, 1, True]
30 [-1, 21] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
31 29 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
32 [-1, 25] 1 0 ultralytics.nn.modules.conv.Concat [1]
33 -1 1 35136 ultralytics.nn.modules.block.DSC3k2 [256, 64, 1, True]
34 22 1 8320 ultralytics.nn.modules.conv.Conv [128, 64, 1, 1]
35 [33, 34] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
36 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
37 [-1, 30] 1 0 ultralytics.nn.modules.conv.Concat [1]
38 -1 1 90752 ultralytics.nn.modules.block.DSC3k2 [192, 128, 1, True]
39 [-1, 21] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
40 38 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
41 [-1, 26] 1 0 ultralytics.nn.modules.conv.Concat [1]
42 -1 1 345344 ultralytics.nn.modules.block.DSC3k2 [384, 256, 1, True]
43 [-1, 23] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
44 [16, 18, 20] 1 273536 ultralytics.nn.modules.block.HyperACE [128, 128, 1, 4, True, True, 0.5, 1, 'both']
45 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
46 44 1 33280 ultralytics.nn.modules.block.DownsampleConv [128]
47 [18, 44] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
48 [16, 45] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
49 [20, 46] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
50 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
51 [-1, 47] 1 0 ultralytics.nn.modules.conv.Concat [1]
52 -1 1 115328 ultralytics.nn.modules.block.DSC3k2 [384, 128, 1, True]
53 [-1, 44] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
54 52 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
55 [-1, 48] 1 0 ultralytics.nn.modules.conv.Concat [1]
56 -1 1 35136 ultralytics.nn.modules.block.DSC3k2 [256, 64, 1, True]
57 45 1 8320 ultralytics.nn.modules.conv.Conv [128, 64, 1, 1]
58 [56, 57] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
59 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
60 [-1, 53] 1 0 ultralytics.nn.modules.conv.Concat [1]
61 -1 1 90752 ultralytics.nn.modules.block.DSC3k2 [192, 128, 1, True]
62 [-1, 44] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
63 61 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
64 [-1, 49] 1 0 ultralytics.nn.modules.conv.Concat [1]
65 -1 1 345344 ultralytics.nn.modules.block.DSC3k2 [384, 256, 1, True]
66 [-1, 46] 1 1 ultralytics.nn.modules.block.FullPAD_Tunnel []
67 [35, 58] 1 56226 ultralytics.nn.AddModules.PSFM.PSFM [64]
68 [39, 62] 1 205634 ultralytics.nn.AddModules.PSFM.PSFM [128]
69 [43, 66] 1 784002 ultralytics.nn.AddModules.PSFM.PSFM [256]
70 [67, 68, 69] 1 432427 ultralytics.nn.modules.head.Detect [9, [64, 128, 256]]
YOLOv13-late-PSFM summary: 1,480 layers, 5,680,767 parameters, 5,680,751 gradients, 14.9 GFLOPs
