【YOLOv8多模态融合改进】在前期、中期、中后期、后期多模态融合中添加P5大目标检测层，完整步骤及代码
前言
主题： YOLOv8 的多模态融合改进中增加P6大目标检测层
方式： 分别在前期融合、中期融合、中-后期融合、后期融合中增加P6多模态融合检测层。
内容： 包含融合方式详解以及完整配置步骤，开箱即用，一键运行。
一、YOLOv8原始模型结构介绍
YOLOv8
原始模型结构如下：
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 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=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024] # YOLOv8n summary: 225 layers, 3157200 parameters, 3157184 gradients, 8.9 GFLOPs
s: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPs
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
x: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs
# YOLOv8.0n 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, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 3-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 5-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 7-P5/32
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 9
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 6], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [512]] # 12
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [[-1, 4], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 15 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 12], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 18 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 9], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 21 (P5/32-large)
- [[15, 18, 21], 1, Detect, [nc]] # Detect(P3, P4, P5)
二、有效特征层对应的检测头类别
2.1 P3/8 - small检测头
-
原始模型中的
P3/8特征层对应的检测头主要用于检测相对较小的目标。其特征图大小相对较大，空间分辨率较高。 -
适合检测尺寸大概在
8x8到16x16像素左右的目标。
2.2 P4/16 - medium检测头
-
这个检测头对应的
P4/16特征层经过了更多的下采样操作，相比P3/8特征图空间分辨率降低，但通道数增加，特征更抽象且有语义信息。 -
它主要用于检测中等大小的目标，尺寸范围大概在
16x16到32x32像素左右。
2.3 P5/32 - large检测头
-
P5/32是经过最多下采样操作得到的特征层，其空间分辨率最低，但语义信息最强、全局感受野最大。 -
该检测头适合检测较大尺寸的目标，一般是尺寸在
32x32像素以上的目标。
2.4 新添加针对大目标的检测头
-
新添加的检测头主要用于检测更大尺寸的目标。尺寸在
64x64像素以上的超大目标。
💡这是因为在目标检测任务中，随着目标尺寸的增大，需要更能关注到整体轮廓的特征图来有效捕捉大目标特征。
三、P6检测层的多模态融合方式
-
前期融合中，在网络输入阶段将多模态数据合并后，增加针对大目标的检测层。
-
中期融合中，在骨干网络中增加针对P6的多模态特征进行融合，以此引出大目标的检测层。
-
中-后期融合中，在颈部的FPN结构中，增加针对P6的多模态特征进行融合，以此引出大目标的检测层。
-
后期融合中，在检测头前增加P6多模态特征进行融合。
四、完整配置步骤
！！！ 私信获取的项目包就已经把相关的多模态输入、训练等改动都已经配好了，只需要新建模型yaml文件，粘贴对应的模型，进行训练即可。 项目包获取及使用教程可参考链接： 《YOLO系列模型的多模态项目》配置使用教程
在什么地方新建，n,s,m,l,x,用哪个版本按自己的需求来即可，和普通的训练步骤一致。
除了模型结构方面的改动，在yaml文件中还传入了一个通道数
ch: 6
表示传入的是双模态，6通道 ，前三个是可见光，后三个是红外。
在default.yaml中也配置了这个参数。
4.1 P6前期融合
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# Parameters
ch: 6
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024] # YOLOv8n summary: 225 layers, 3157200 parameters, 3157184 gradients, 8.9 GFLOPs
s: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPs
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
x: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs
# YOLOv8.0n backbone
backbone:
# [from, repeats, module, args]
- [-1, 1, MF, [64]] # 0
- [-1, 1, Conv, [64, 3, 2]] # 1-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 2-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 4-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 6-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 8-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 10-P6/64
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 12
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 9], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 15
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 7], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 18 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 5], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 21 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 18], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 24 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 15], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [768]] # 27 (P4/16-medium)
- [-1, 1, Conv, [768, 3, 2]]
- [[-1, 12], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 30 (P5/32-large)
- [[21, 24, 27, 30], 1, Detect, [nc]] # Detect(P3, P4, P5)
4.2 P6中期融合
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# Parameters
ch: 6
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024] # YOLOv8n summary: 225 layers, 3157200 parameters, 3157184 gradients, 8.9 GFLOPs
s: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPs
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
x: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs
# YOLOv8.0n 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]] # 4-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 6-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 8-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 10-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 12-P6/64
- [-1, 3, C2f, [1024, True]]
- [2, 1, Conv, [64, 3, 2]] # 14-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 15-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 17-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 19-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 21-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 23-P6/64
- [-1, 3, C2f, [1024, True]]
- [[7, 18], 1, Concat, [1]] # 25 cat backbone P3
- [[9, 20], 1, Concat, [1]] # 26 cat backbone P4
- [[11, 22], 1, Concat, [1]] # 27 cat backbone P5
- [[13, 24], 1, Concat, [1]] # 28 cat backbone P6
- [-1, 1, SPPF, [1024, 5]] # 29
# YOLOv8.0n head
head:
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 27], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 32
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 26], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 35 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 25], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 38 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 35], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 41 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 32], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [768]] # 44 (P4/16-medium)
- [-1, 1, Conv, [768, 3, 2]]
- [[-1, 29], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 47 (P5/32-large)
- [[38, 41, 44, 47], 1, Detect, [nc]] # Detect(P3, P4, P5)
4.3 P6中-后期融合
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# Parameters
ch: 6
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024] # YOLOv8n summary: 225 layers, 3157200 parameters, 3157184 gradients, 8.9 GFLOPs
s: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPs
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
x: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs
# YOLOv8.0n 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]] # 4-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 6-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 8-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 10-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 12-P6/64
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 14
- [2, 1, Conv, [64, 3, 2]] # 15-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 16-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 18-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 20-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 22-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 24-P6/64
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 26
# YOLOv8.0n head
head:
- [14, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 11], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 29
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 9], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 32 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 7], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 35 (P3/8-small)
- [26, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 23], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 38
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 21], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 41 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 19], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 44 (P3/8-small)
- [ [ 14, 26 ], 1, Concat, [ 1 ] ] # cat head P3 45
- [ [ 29, 38 ], 1, Concat, [ 1 ] ] # cat head P4 46
- [ [ 32, 41 ], 1, Concat, [ 1 ] ] # cat head P5 47
- [ [ 35, 44 ], 1, Concat, [ 1 ] ] # cat head P6 48
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 47], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 51 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 46], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [768]] # 54 (P4/16-medium)
- [-1, 1, Conv, [768, 3, 2]]
- [[-1, 45], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 57 (P5/32-large)
- [[48, 51, 54, 57], 1, Detect, [nc]] # Detect(P3, P4, P5)
4.4 P6后期融合
# Ultralytics YOLO 🚀, AGPL-3.0 license
# YOLOv8 object detection model with P3-P5 outputs. For Usage examples see https://docs.ultralytics.com/tasks/detect
# Parameters
ch: 6
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n.yaml' will call yolov8.yaml with scale 'n'
# [depth, width, max_channels]
n: [0.33, 0.25, 1024] # YOLOv8n summary: 225 layers, 3157200 parameters, 3157184 gradients, 8.9 GFLOPs
s: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPs
m: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPs
l: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPs
x: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOPs
# YOLOv8.0n 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]] # 4-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 6-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 8-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 10-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 12-P6/64
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 14
- [2, 1, Conv, [64, 3, 2]] # 15-P1/2
- [-1, 1, Conv, [128, 3, 2]] # 16-P2/4
- [-1, 3, C2f, [128, True]]
- [-1, 1, Conv, [256, 3, 2]] # 18-P3/8
- [-1, 6, C2f, [256, True]]
- [-1, 1, Conv, [512, 3, 2]] # 20-P4/16
- [-1, 6, C2f, [512, True]]
- [-1, 1, Conv, [768, 3, 2]] # 22-P5/32
- [-1, 3, C2f, [768, True]]
- [-1, 1, Conv, [1024, 3, 2]] # 24-P6/64
- [-1, 3, C2f, [1024, True]]
- [-1, 1, SPPF, [1024, 5]] # 26
# YOLOv8.0n head
head:
- [14, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 11], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 29
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 9], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 32 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 7], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 35 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 32], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 38 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 29], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [768]] # 41 (P4/16-medium)
- [-1, 1, Conv, [768, 3, 2]]
- [[-1, 14], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 44 (P5/32-large)
- [26, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 23], 1, Concat, [1]] # cat backbone P4
- [-1, 3, C2f, [768]] # 47
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 21], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [512]] # 50 (P3/8-small)
- [-1, 1, nn.Upsample, [None, 2, 'nearest']]
- [[-1, 19], 1, Concat, [1]] # cat backbone P3
- [-1, 3, C2f, [256]] # 53 (P3/8-small)
- [-1, 1, Conv, [256, 3, 2]]
- [[-1, 50], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [512]] # 56 (P4/16-medium)
- [-1, 1, Conv, [512, 3, 2]]
- [[-1, 47], 1, Concat, [1]] # cat head P4
- [-1, 3, C2f, [768]] # 59 (P4/16-medium)
- [-1, 1, Conv, [768, 3, 2]]
- [[-1, 26], 1, Concat, [1]] # cat head P5
- [-1, 3, C2f, [1024]] # 62 (P5/32-large)
- [ [ 35, 53 ], 1, Concat, [ 1 ] ] # cat head P3 63
- [ [ 38, 56 ], 1, Concat, [ 1 ] ] # cat head P4 64
- [ [ 41, 59 ], 1, Concat, [ 1 ] ] # cat head P5 65
- [ [ 44, 62 ], 1, Concat, [ 1 ] ] # cat head P6 66
- [[63, 64, 65, 66], 1, Detect, [nc]] # Detect(P3, P4, P5)
五、成功运行结果
前期融合结果： 可以看到输入的通道数为6，表明可见光图像和红外图像均输入到了模型中进行融合训练。
YOLOv8-early-p6 summary: 340 layers, 4,424,844 parameters, 4,424,828 gradients, 7.8 GFLOPs
from n params module arguments
0 -1 1 472 ultralytics.nn.AddModules.multimodal.MF [6, 16]
1 -1 1 2336 ultralytics.nn.modules.conv.Conv [16, 16, 3, 2]
2 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
3 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
4 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
5 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
6 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
7 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
8 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
9 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
10 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
11 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
12 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
13 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
14 [-1, 9] 1 0 ultralytics.nn.modules.conv.Concat [1]
15 -1 1 308352 ultralytics.nn.modules.block.C2f [448, 192, 1]
16 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
17 [-1, 7] 1 0 ultralytics.nn.modules.conv.Concat [1]
18 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
19 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
20 [-1, 5] 1 0 ultralytics.nn.modules.conv.Concat [1]
21 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
22 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
23 [-1, 18] 1 0 ultralytics.nn.modules.conv.Concat [1]
24 -1 1 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
25 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
26 [-1, 15] 1 0 ultralytics.nn.modules.conv.Concat [1]
27 -1 1 283776 ultralytics.nn.modules.block.C2f [320, 192, 1]
28 -1 1 332160 ultralytics.nn.modules.conv.Conv [192, 192, 3, 2]
29 [-1, 12] 1 0 ultralytics.nn.modules.conv.Concat [1]
30 -1 1 509440 ultralytics.nn.modules.block.C2f [448, 256, 1]
31 [21, 24, 27, 30] 1 602260 ultralytics.nn.modules.head.Detect [1, [64, 128, 192, 256]]
YOLOv8-early-p6 summary: 340 layers, 4,424,844 parameters, 4,424,828 gradients, 7.8 GFLOPs
中期融合结果：
YOLOv8-mid-p6 summary: 441 layers, 6,445,748 parameters, 6,445,732 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 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
5 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
6 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
7 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
8 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
9 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
10 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
11 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
12 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
13 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
14 2 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2]
15 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
16 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
17 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
18 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
19 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
20 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
21 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
22 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
23 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
24 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
25 [7, 18] 1 0 ultralytics.nn.modules.conv.Concat [1]
26 [9, 20] 1 0 ultralytics.nn.modules.conv.Concat [1]
27 [11, 22] 1 0 ultralytics.nn.modules.conv.Concat [1]
28 [13, 24] 1 0 ultralytics.nn.modules.conv.Concat [1]
29 -1 1 394240 ultralytics.nn.modules.block.SPPF [512, 256, 5]
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 345216 ultralytics.nn.modules.block.C2f [640, 192, 1]
33 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
34 [-1, 26] 1 0 ultralytics.nn.modules.conv.Concat [1]
35 -1 1 156416 ultralytics.nn.modules.block.C2f [448, 128, 1]
36 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
37 [-1, 25] 1 0 ultralytics.nn.modules.conv.Concat [1]
38 -1 1 41344 ultralytics.nn.modules.block.C2f [256, 64, 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 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
42 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
43 [-1, 32] 1 0 ultralytics.nn.modules.conv.Concat [1]
44 -1 1 283776 ultralytics.nn.modules.block.C2f [320, 192, 1]
45 -1 1 332160 ultralytics.nn.modules.conv.Conv [192, 192, 3, 2]
46 [-1, 29] 1 0 ultralytics.nn.modules.conv.Concat [1]
47 -1 1 509440 ultralytics.nn.modules.block.C2f [448, 256, 1]
48 [38, 41, 44, 47] 1 602260 ultralytics.nn.modules.head.Detect [1, [64, 128, 192, 256]]
YOLOv8-mid-p6 summary: 441 layers, 6,445,748 parameters, 6,445,732 gradients, 10.3 GFLOPs
中-后期融合结果：
YOLOv8-mid-to-late-p6 summary: 500 layers, 7,104,628 parameters, 7,104,612 gradients, 12.4 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 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
5 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
6 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
7 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
8 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
9 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
10 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
11 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
12 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
13 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
14 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
15 2 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2]
16 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
17 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
18 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
19 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
20 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
21 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
22 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
23 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
24 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
25 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
26 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
27 14 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
28 [-1, 11] 1 0 ultralytics.nn.modules.conv.Concat [1]
29 -1 1 308352 ultralytics.nn.modules.block.C2f [448, 192, 1]
30 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
31 [-1, 9] 1 0 ultralytics.nn.modules.conv.Concat [1]
32 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
33 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
34 [-1, 7] 1 0 ultralytics.nn.modules.conv.Concat [1]
35 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
36 26 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
37 [-1, 23] 1 0 ultralytics.nn.modules.conv.Concat [1]
38 -1 1 308352 ultralytics.nn.modules.block.C2f [448, 192, 1]
39 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
40 [-1, 21] 1 0 ultralytics.nn.modules.conv.Concat [1]
41 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
42 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
43 [-1, 19] 1 0 ultralytics.nn.modules.conv.Concat [1]
44 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
45 [14, 26] 1 0 ultralytics.nn.modules.conv.Concat [1]
46 [29, 38] 1 0 ultralytics.nn.modules.conv.Concat [1]
47 [32, 41] 1 0 ultralytics.nn.modules.conv.Concat [1]
48 [35, 44] 1 0 ultralytics.nn.modules.conv.Concat [1]
49 -1 1 73856 ultralytics.nn.modules.conv.Conv [128, 64, 3, 2]
50 [-1, 47] 1 0 ultralytics.nn.modules.conv.Concat [1]
51 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
52 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
53 [-1, 46] 1 0 ultralytics.nn.modules.conv.Concat [1]
54 -1 1 320640 ultralytics.nn.modules.block.C2f [512, 192, 1]
55 -1 1 332160 ultralytics.nn.modules.conv.Conv [192, 192, 3, 2]
56 [-1, 45] 1 0 ultralytics.nn.modules.conv.Concat [1]
57 -1 1 574976 ultralytics.nn.modules.block.C2f [704, 256, 1]
58 [48, 51, 54, 57] 1 742228 ultralytics.nn.modules.head.Detect [1, [128, 128, 192, 256]]
YOLOv8-mid-to-late-p6 summary: 500 layers, 7,104,628 parameters, 7,104,612 gradients, 12.4 GFLOPs
后期融合结果：
YOLOv8-late-p6 summary: 557 layers, 8,794,548 parameters, 8,794,532 gradients, 13.7 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 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
5 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
6 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
7 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
8 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
9 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
10 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
11 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
12 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
13 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
14 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
15 2 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2]
16 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2]
17 -1 1 7360 ultralytics.nn.modules.block.C2f [32, 32, 1, True]
18 -1 1 18560 ultralytics.nn.modules.conv.Conv [32, 64, 3, 2]
19 -1 2 49664 ultralytics.nn.modules.block.C2f [64, 64, 2, True]
20 -1 1 73984 ultralytics.nn.modules.conv.Conv [64, 128, 3, 2]
21 -1 2 197632 ultralytics.nn.modules.block.C2f [128, 128, 2, True]
22 -1 1 221568 ultralytics.nn.modules.conv.Conv [128, 192, 3, 2]
23 -1 1 259200 ultralytics.nn.modules.block.C2f [192, 192, 1, True]
24 -1 1 442880 ultralytics.nn.modules.conv.Conv [192, 256, 3, 2]
25 -1 1 460288 ultralytics.nn.modules.block.C2f [256, 256, 1, True]
26 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5]
27 14 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
28 [-1, 11] 1 0 ultralytics.nn.modules.conv.Concat [1]
29 -1 1 308352 ultralytics.nn.modules.block.C2f [448, 192, 1]
30 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
31 [-1, 9] 1 0 ultralytics.nn.modules.conv.Concat [1]
32 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
33 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
34 [-1, 7] 1 0 ultralytics.nn.modules.conv.Concat [1]
35 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
36 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
37 [-1, 32] 1 0 ultralytics.nn.modules.conv.Concat [1]
38 -1 1 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
39 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
40 [-1, 29] 1 0 ultralytics.nn.modules.conv.Concat [1]
41 -1 1 283776 ultralytics.nn.modules.block.C2f [320, 192, 1]
42 -1 1 332160 ultralytics.nn.modules.conv.Conv [192, 192, 3, 2]
43 [-1, 14] 1 0 ultralytics.nn.modules.conv.Concat [1]
44 -1 1 509440 ultralytics.nn.modules.block.C2f [448, 256, 1]
45 26 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
46 [-1, 23] 1 0 ultralytics.nn.modules.conv.Concat [1]
47 -1 1 308352 ultralytics.nn.modules.block.C2f [448, 192, 1]
48 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
49 [-1, 21] 1 0 ultralytics.nn.modules.conv.Concat [1]
50 -1 1 140032 ultralytics.nn.modules.block.C2f [320, 128, 1]
51 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
52 [-1, 19] 1 0 ultralytics.nn.modules.conv.Concat [1]
53 -1 1 37248 ultralytics.nn.modules.block.C2f [192, 64, 1]
54 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2]
55 [-1, 50] 1 0 ultralytics.nn.modules.conv.Concat [1]
56 -1 1 123648 ultralytics.nn.modules.block.C2f [192, 128, 1]
57 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2]
58 [-1, 47] 1 0 ultralytics.nn.modules.conv.Concat [1]
59 -1 1 283776 ultralytics.nn.modules.block.C2f [320, 192, 1]
60 -1 1 332160 ultralytics.nn.modules.conv.Conv [192, 192, 3, 2]
61 [-1, 26] 1 0 ultralytics.nn.modules.conv.Concat [1]
62 -1 1 509440 ultralytics.nn.modules.block.C2f [448, 256, 1]
63 [35, 53] 1 0 ultralytics.nn.modules.conv.Concat [1]
64 [38, 56] 1 0 ultralytics.nn.modules.conv.Concat [1]
65 [41, 59] 1 0 ultralytics.nn.modules.conv.Concat [1]
66 [44, 62] 1 0 ultralytics.nn.modules.conv.Concat [1]
67 [63, 64, 65, 66] 1 1154068 ultralytics.nn.modules.head.Detect [1, [128, 256, 384, 512]]
YOLOv8-late-p6 summary: 557 layers, 8,794,548 parameters, 8,794,532 gradients, 13.7 GFLOPs