RT-DETR改进策略【卷积层】| AAAI 2025 风车状卷积PConv，实现感受野的高效扩张
一、本文介绍
本文记录的是 利用风车卷积改进RT-DETR的目标检测网络模型。
在红外小目标检测任务中，传统卷积方式
难以捕捉目标像素的空间特征
，影响检测性能，因此需要更适配的卷积方式提升特征提取能力。但
不同尺度的红外小目标对特征提取需求有差异
，为了更好地满足这些需求，本文利用
风车卷积PConv
模块改进
RT-DETR
，使模型能够
更精准地对齐红外小目标像素的高斯空间分布，在增强底层特征提取的同时显著扩大感受野
，使网络更好地适应不同尺度红外小目标的检测需求。
二、风车卷积介绍
Pinwheel-shaped Convolution and Scale-based Dynamic Loss for Infrared Small Target Detection
2.1 设计出发点
传统基于卷积神经网络的红外小目标检测方法多采用标准卷积，忽略了红外小目标像素分布的空间特性。分析红外小目标的3D灰度分布可知其具有高斯特征，而标准卷积无法很好地契合这一特性，影响了底层特征提取效果。为提升CNN对红外小目标底层特征的分析能力，提出了PConv模块。
2.2 结构原理
PConv模块采用非对称填充方式，为图像不同区域创建水平和垂直卷积核 。
模块内部通过多层卷积操作，对输入张量 X ( h 1 , w 1 , c 1 ) X^{(h_{1}, w_{1}, c_{1})} X ( h 1 , w 1 , c 1 ) 进行处理。第一层进行并行卷积，利用公式计算得到多个中间结果，如 X 1 ( h ′ , w ′ , c ′ ) = S i L U ( B N ( X P ( 1 , 0 , 0 , 3 ) ( h 1 , w 1 , c 1 ) ⊗ W 1 ( 1 , 3 , c ′ ) ) ) X_{1}^{\left(h', w', c'\right)}=SiLU\left(BN\left(X_{P(1,0,0,3)}^{\left(h_{1}, w_{1}, c_{1}\right)} \otimes W_{1}^{\left(1,3, c'\right)}\right)\right) X 1 ( h ′ , w ′ , c ′ ) = S i LU ( BN ( X P ( 1 , 0 , 0 , 3 ) ( h 1 , w 1 , c 1 ) ⊗ W 1 ( 1 , 3 , c ′ ) ) ) 等。这些结果经过拼接后，再由特定卷积核归一化处理，最终得到输出特征图。其输出特征图的尺寸与输入特征图尺寸存在特定关系，如 h ′ = h 1 s + 1 h'=\frac{h_{1}}{s}+1 h ′ = s h 1 + 1 等。
此外，PConv模块还利用分组卷积，在增加感受野的同时减少参数数量。
2.3优势
- 增强特征提取，PConv模块更好地对齐了红外小目标的像素高斯空间分布，相比标准卷积，能更有效地提取底层特征，通过可视化结果可知其增强了目标与背景的对比度，抑制了杂波信号。
- 显著扩大感受野，PConv模块的感受野类似高斯分布，向外逐渐减弱 。以 k = 3 k=3 k = 3 的PConv为例，其感受野为25，相比标准卷积有大幅提升。
- 参数增加少，在不同卷积核设置下，与标准卷积相比，PConv在增加感受野的同时，参数增加幅度较小，如 k = 3 k=3 k = 3 时，相比 3 × 3 3×3 3 × 3 标准卷积，PConv增加111%的参数，却使感受野扩大178%，实现了高效的感受野扩展和参数利用。
论文： https://arxiv.org/pdf/2412.16986v1
源码： https://github.com/JN-Yang/PConv-SDloss-Data
三、风车卷积的实现代码
风车卷积模块
的实现代码如下：
import torch
import torch.nn as nn
import torch.nn.functional as F
from ultralytics.nn.modules.conv import LightConv
"""风车型卷积，使用了padding再各个方向上实现方向敏感性，增加了参数量"""
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 PConv(nn.Module):
''' Pinwheel-shaped Convolution using the Asymmetric Padding method. '''
def __init__(self, c1, c2, k, s):
super().__init__()
# self.k = k
p = [(k, 0, 1, 0), (0, k, 0, 1), (0, 1, k, 0), (1, 0, 0, k)]
self.pad = [nn.ZeroPad2d(padding=(p[g])) for g in range(4)]
self.cw = Conv(c1, c2 // 4, (1, k), s=s, p=0)
self.ch = Conv(c1, c2 // 4, (k, 1), s=s, p=0)
self.cat = Conv(c2, c2, 2, s=1, p=0)
def forward(self, x):
yw0 = self.cw(self.pad[0](x))
yw1 = self.cw(self.pad[1](x))
yh0 = self.ch(self.pad[2](x))
yh1 = self.ch(self.pad[3](x))
return self.cat(torch.cat([yw0, yw1, yh0, yh1], dim=1))
class HGBlock_APConv(nn.Module):
"""
HG_Block of PPHGNetV2 with 2 convolutions and LightConv.
https://github.com/PaddlePaddle/PaddleDetection/blob/develop/ppdet/modeling/backbones/hgnet_v2.py
"""
def __init__(self, c1, cm, c2, k=3, n=6, lightconv=False, shortcut=False, act=nn.ReLU()):
"""Initializes a CSP Bottleneck with 1 convolution using specified input and output channels."""
super().__init__()
block = LightConv if lightconv else Conv
self.m = nn.ModuleList(block(c1 if i == 0 else cm, cm, k=k, act=act) for i in range(n))
self.sc = Conv(c1 + n * cm, c2 // 2, 1, 1, act=act) # squeeze conv
self.ec = Conv(c2 // 2, c2, 1, 1, act=act) # excitation conv
self.add = shortcut and c1 == c2
self.cv = PConv(c1, c2, 3, 1)
def forward(self, x):
"""Forward pass of a PPHGNetV2 backbone layer."""
y = [x]
y.extend(m(y[-1]) for m in self.m)
y = self.cv(self.ec(self.sc(torch.cat(y, 1))))
return y + x if self.add else y
class ResNetBlock(nn.Module):
"""ResNet block with standard convolution layers."""
def __init__(self, c1, c2, s=1, e=4):
"""Initialize convolution with given parameters."""
super().__init__()
c3 = e * c2
self.cv1 = Conv(c1, c2, k=1, s=1, act=True)
self.cv2 = Conv(c2, c2, k=3, s=s, p=1, act=True)
self.cv3 = Conv(c2, c3, k=1, act=False)
self.cv4 = PConv(c2, c2, 3, 1)
self.shortcut = nn.Sequential(Conv(c1, c3, k=1, s=s, act=False)) if s != 1 or c1 != c3 else nn.Identity()
def forward(self, x):
"""Forward pass through the ResNet block."""
return F.relu(self.cv3(self.cv4(self.cv2(self.cv1(x)))) + self.shortcut(x))
class ResNetLayer_APConv(nn.Module):
"""ResNet layer with multiple ResNet blocks."""
def __init__(self, c1, c2, s=1, is_first=False, n=1, e=4):
"""Initializes the ResNetLayer given arguments."""
super().__init__()
self.is_first = is_first
if self.is_first:
self.layer = nn.Sequential(
Conv(c1, c2, k=7, s=2, p=3, act=True), nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
)
else:
blocks = [ResNetBlock(c1, c2, s, e=e)]
blocks.extend([ResNetBlock(e * c2, c2, 1, e=e) for _ in range(n - 1)])
self.layer = nn.Sequential(*blocks)
def forward(self, x):
"""Forward pass through the ResNet layer."""
return self.layer(x)
if __name__ == "__main__":
x = torch.randn(1, 32, 64, 64).cuda()
xm = torch.randn(1, 1, 320, 320).cuda()
model = PConv(c1=32,c2=32,k=3,s=1).cuda()
#model = Conv(c1=32, c2=32, k=3, s=1).cuda()
y = model(x)
print(y.size())
print("最大内存占用:", torch.cuda.max_memory_allocated() // 1024 // 1024, "MB")
# | 模块 | 参数量 | FLOPs | 特征提取能力 |
# |--------|--------|---------|--------------------|
# | Conv | 3.7K | 2.4G | 各向同性特征 |
# | PConv | 25.6K | 3.2G | 方向敏感特征 |
四、创新模块
4.1 改进点⭐
模块改进方法
：在
HGBlock
中加入
风车卷积
（
第五节讲解添加步骤
）。
风车卷积
添加后如下：
class HGBlock_APConv(nn.Module):
"""
HG_Block of PPHGNetV2 with 2 convolutions and LightConv.
https://github.com/PaddlePaddle/PaddleDetection/blob/develop/ppdet/modeling/backbones/hgnet_v2.py
"""
def __init__(self, c1, cm, c2, k=3, n=6, lightconv=False, shortcut=False, act=nn.ReLU()):
"""Initializes a CSP Bottleneck with 1 convolution using specified input and output channels."""
super().__init__()
block = LightConv if lightconv else Conv
self.m = nn.ModuleList(block(c1 if i == 0 else cm, cm, k=k, act=act) for i in range(n))
self.sc = Conv(c1 + n * cm, c2 // 2, 1, 1, act=act) # squeeze conv
self.ec = Conv(c2 // 2, c2, 1, 1, act=act) # excitation conv
self.add = shortcut and c1 == c2
self.cv = PConv(c1, c2, 3, 1)
def forward(self, x):
"""Forward pass of a PPHGNetV2 backbone layer."""
y = [x]
y.extend(m(y[-1]) for m in self.m)
y = self.cv(self.ec(self.sc(torch.cat(y, 1))))
return y + x if self.add else y
4.2 改进点2⭐
模块改进方法
：基于
风车卷积模块
的
ResNetLayer
（
第五节讲解添加步骤
）。
改进代码如下：
首先添加
风车卷积
改进
ResNetBlock
模块。
class ResNetBlock(nn.Module):
"""ResNet block with standard convolution layers."""
def __init__(self, c1, c2, s=1, e=4):
"""Initialize convolution with given parameters."""
super().__init__()
c3 = e * c2
self.cv1 = Conv(c1, c2, k=1, s=1, act=True)
self.cv2 = Conv(c2, c2, k=3, s=s, p=1, act=True)
self.cv3 = Conv(c2, c3, k=1, act=False)
self.cv4 = PConv(c2, c2, 3, 1)
self.shortcut = nn.Sequential(Conv(c1, c3, k=1, s=s, act=False)) if s != 1 or c1 != c3 else nn.Identity()
def forward(self, x):
"""Forward pass through the ResNet block."""
return F.relu(self.cv3(self.cv4(self.cv2(self.cv1(x)))) + self.shortcut(x))
然后将
ResNetLayer
重命名为
ResNetLayer_APConv
class ResNetLayer_APConv(nn.Module):
"""ResNet layer with multiple ResNet blocks."""
def __init__(self, c1, c2, s=1, is_first=False, n=1, e=4):
"""Initializes the ResNetLayer given arguments."""
super().__init__()
self.is_first = is_first
if self.is_first:
self.layer = nn.Sequential(
Conv(c1, c2, k=7, s=2, p=3, act=True), nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
)
else:
blocks = [ResNetBlock(c1, c2, s, e=e)]
blocks.extend([ResNetBlock(e * c2, c2, 1, e=e) for _ in range(n - 1)])
self.layer = nn.Sequential(*blocks)
def forward(self, x):
"""Forward pass through the ResNet layer."""
return self.layer(x)
注意❗：在
第五小节
中需要声明的模块名称为：
HGBlock_APConv
和
ResNetLayer_APConv
。
五、添加步骤
5.1 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
APConv.py
，将
第三节
中的代码粘贴到此处
5.2 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 .APConv import *
5.3 修改三
在
ultralytics/nn/modules/tasks.py
文件中，需要在指定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
HGBlock_APConv
和
ResNetLayer_APConv
模块
在
parse_model函数
的如下添加模块：
六、yaml模型文件
6.1 模型改进版本⭐
此处以
ultralytics/cfg/models/rt-detr/rtdetr-l.yaml
为例，在同目录下创建一个用于自己数据集训练的模型文件
rtdetr-l-HGBlock_APConv.yaml
。
将
rtdetr-l.yaml
中的内容复制到
rtdetr-l-HGBlock_APConv.yaml
文件下，修改
nc
数量等于自己数据中目标的数量。
📌 模型的修改方法是将
骨干网络
中的
HGBlock模块
替换成
HGBlock_APConv模块
。
# Ultralytics YOLO 🚀, AGPL-3.0 license
# RT-DETR-l object detection model with P3-P5 outputs. For details see https://docs.ultralytics.com/models/rtdetr
# Parameters
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n-cls.yaml' will call yolov8-cls.yaml with scale 'n'
# [depth, width, max_channels]
l: [1.00, 1.00, 1024]
backbone:
# [from, repeats, module, args]
- [-1, 1, HGStem, [32, 48]] # 0-P2/4
- [-1, 6, HGBlock, [48, 128, 3]] # stage 1
- [-1, 1, DWConv, [128, 3, 2, 1, False]] # 2-P3/8
- [-1, 6, HGBlock, [96, 512, 3]] # stage 2
- [-1, 1, DWConv, [512, 3, 2, 1, False]] # 4-P4/16
- [-1, 6, HGBlock_APConv, [192, 512, 5, True, False]] # cm, c2, k, light, shortcut
- [-1, 6, HGBlock_APConv, [192, 512, 5, True, True]]
- [-1, 6, HGBlock_APConv, [192, 512, 5, True, True]] # stage 3
- [-1, 1, DWConv, [1024, 3, 2, 1, False]] # 8-P5/32
- [-1, 6, HGBlock_APConv, [384, 1024, 5, True, False]] # stage 4
head:
- [-1, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 10 input_proj.2
- [-1, 1, AIFI, [1024, 8]]
- [-1, 1, Conv, [256, 1, 1]] # 12, Y5, lateral_convs.0
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [7, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 14 input_proj.1
- [[-2, -1], 1, Concat, [1]]
- [-1, 3, RepC3, [256]] # 16, fpn_blocks.0
- [-1, 1, Conv, [256, 1, 1]] # 17, Y4, lateral_convs.1
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [3, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 19 input_proj.0
- [[-2, -1], 1, Concat, [1]] # cat backbone P4
- [-1, 3, RepC3, [256]] # X3 (21), fpn_blocks.1
- [-1, 1, Conv, [256, 3, 2]] # 22, downsample_convs.0
- [[-1, 17], 1, Concat, [1]] # cat Y4
- [-1, 3, RepC3, [256]] # F4 (24), pan_blocks.0
- [-1, 1, Conv, [256, 3, 2]] # 25, downsample_convs.1
- [[-1, 12], 1, Concat, [1]] # cat Y5
- [-1, 3, RepC3, [256]] # F5 (27), pan_blocks.1
- [[21, 24, 27], 1, RTDETRDecoder, [nc]] # Detect(P3, P4, P5)
6.2 模型改进版本2⭐
此处以
ultralytics/cfg/models/rt-detr/rtdetr-resnet50.yaml
为例，在同目录下创建一个用于自己数据集训练的模型文件
rtdetr-ResNetLayer_APConv.yaml
。
将
rtdetr-resnet50.yaml
中的内容复制到
rtdetr-ResNetLayer_APConv.yaml
文件下，修改
nc
数量等于自己数据中目标的数量。
📌 模型的修改方法是将
骨干网络
中的所有
ResNetLayer模块
替换成
ResNetLayer_APConv模块
。
# Ultralytics YOLO 🚀, AGPL-3.0 license
# RT-DETR-ResNet50 object detection model with P3-P5 outputs.
# Parameters
nc: 1 # number of classes
scales: # model compound scaling constants, i.e. 'model=yolov8n-cls.yaml' will call yolov8-cls.yaml with scale 'n'
# [depth, width, max_channels]
l: [1.00, 1.00, 1024]
backbone:
# [from, repeats, module, args]
- [-1, 1, ResNetLayer_APConv, [3, 64, 1, True, 1]] # 0
- [-1, 1, ResNetLayer_APConv, [64, 64, 1, False, 3]] # 1
- [-1, 1, ResNetLayer_APConv, [256, 128, 2, False, 4]] # 2
- [-1, 1, ResNetLayer_APConv, [512, 256, 2, False, 6]] # 3
- [-1, 1, ResNetLayer_APConv, [1024, 512, 2, False, 3]] # 4
head:
- [-1, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 5
- [-1, 1, AIFI, [1024, 8]]
- [-1, 1, Conv, [256, 1, 1]] # 7
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [3, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 9
- [[-2, -1], 1, Concat, [1]]
- [-1, 3, RepC3, [256]] # 11
- [-1, 1, Conv, [256, 1, 1]] # 12
- [-1, 1, nn.Upsample, [None, 2, "nearest"]]
- [2, 1, Conv, [256, 1, 1, None, 1, 1, False]] # 14
- [[-2, -1], 1, Concat, [1]] # cat backbone P4
- [-1, 3, RepC3, [256]] # X3 (16), fpn_blocks.1
- [-1, 1, Conv, [256, 3, 2]] # 17, downsample_convs.0
- [[-1, 12], 1, Concat, [1]] # cat Y4
- [-1, 3, RepC3, [256]] # F4 (19), pan_blocks.0
- [-1, 1, Conv, [256, 3, 2]] # 20, downsample_convs.1
- [[-1, 7], 1, Concat, [1]] # cat Y5
- [-1, 3, RepC3, [256]] # F5 (22), pan_blocks.1
- [[16, 19, 22], 1, RTDETRDecoder, [nc]] # Detect(P3, P4, P5)
七、成功运行结果
打印网络模型可以看到
HGBlock_APConv
和
ResNetLayer_APConv
已经加入到模型中，并可以进行训练了。
rtdetr-l-HGBlock_APConv ：
rtdetr-l-HGBlock_APConv summary: 723 layers, 36,052,163 parameters, 36,052,163 gradients, 118.7 GFLOPs
from n params module arguments
0 -1 1 25248 ultralytics.nn.modules.block.HGStem [3, 32, 48]
1 -1 6 155072 ultralytics.nn.modules.block.HGBlock [48, 48, 128, 3, 6]
2 -1 1 1408 ultralytics.nn.modules.conv.DWConv [128, 128, 3, 2, 1, False]
3 -1 6 839296 ultralytics.nn.modules.block.HGBlock [128, 96, 512, 3, 6]
4 -1 1 5632 ultralytics.nn.modules.conv.DWConv [512, 512, 3, 2, 1, False]
5 -1 6 2317952 ultralytics.nn.AddModules.APConv.HGBlock_APConv[512, 192, 512, 5, 6, True, False]
6 -1 6 2317952 ultralytics.nn.AddModules.APConv.HGBlock_APConv[512, 192, 512, 5, 6, True, True]
7 -1 6 2317952 ultralytics.nn.AddModules.APConv.HGBlock_APConv[512, 192, 512, 5, 6, True, True]
8 -1 1 11264 ultralytics.nn.modules.conv.DWConv [512, 1024, 3, 2, 1, False]
9 -1 6 9198848 ultralytics.nn.AddModules.APConv.HGBlock_APConv[1024, 384, 1024, 5, 6, True, False]
10 -1 1 262656 ultralytics.nn.modules.conv.Conv [1024, 256, 1, 1, None, 1, 1, False]
11 -1 1 789760 ultralytics.nn.modules.transformer.AIFI [256, 1024, 8]
12 -1 1 66048 ultralytics.nn.modules.conv.Conv [256, 256, 1, 1]
13 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
14 7 1 131584 ultralytics.nn.modules.conv.Conv [512, 256, 1, 1, None, 1, 1, False]
15 [-2, -1] 1 0 ultralytics.nn.modules.conv.Concat [1]
16 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
17 -1 1 66048 ultralytics.nn.modules.conv.Conv [256, 256, 1, 1]
18 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
19 3 1 131584 ultralytics.nn.modules.conv.Conv [512, 256, 1, 1, None, 1, 1, False]
20 [-2, -1] 1 0 ultralytics.nn.modules.conv.Concat [1]
21 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
22 -1 1 590336 ultralytics.nn.modules.conv.Conv [256, 256, 3, 2]
23 [-1, 17] 1 0 ultralytics.nn.modules.conv.Concat [1]
24 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
25 -1 1 590336 ultralytics.nn.modules.conv.Conv [256, 256, 3, 2]
26 [-1, 12] 1 0 ultralytics.nn.modules.conv.Concat [1]
27 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
28 [21, 24, 27] 1 7303907 ultralytics.nn.modules.head.RTDETRDecoder [1, [256, 256, 256]]
rtdetr-l-HGBlock_APConv summary: 723 layers, 36,052,163 parameters, 36,052,163 gradients, 118.7 GFLOPs
rtdetr-ResNetLayer_APConv ：
rtdetr-ResNetLayer_APConv summary: 753 layers, 49,690,211 parameters, 49,690,211 gradients, 154.6 GFLOPs
from n params module arguments
0 -1 1 9536 ultralytics.nn.AddModules.APConv.ResNetLayer_APConv[3, 64, 1, True, 1]
1 -1 1 283968 ultralytics.nn.AddModules.APConv.ResNetLayer_APConv[64, 64, 1, False, 3]
2 -1 1 1581568 ultralytics.nn.AddModules.APConv.ResNetLayer_APConv[256, 128, 2, False, 4]
3 -1 1 9265664 ultralytics.nn.AddModules.APConv.ResNetLayer_APConv[512, 256, 2, False, 6]
4 -1 1 19294720 ultralytics.nn.AddModules.APConv.ResNetLayer_APConv[1024, 512, 2, False, 3]
5 -1 1 524800 ultralytics.nn.modules.conv.Conv [2048, 256, 1, 1, None, 1, 1, False]
6 -1 1 789760 ultralytics.nn.modules.transformer.AIFI [256, 1024, 8]
7 -1 1 66048 ultralytics.nn.modules.conv.Conv [256, 256, 1, 1]
8 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
9 3 1 262656 ultralytics.nn.modules.conv.Conv [1024, 256, 1, 1, None, 1, 1, False]
10 [-2, -1] 1 0 ultralytics.nn.modules.conv.Concat [1]
11 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
12 -1 1 66048 ultralytics.nn.modules.conv.Conv [256, 256, 1, 1]
13 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest']
14 2 1 131584 ultralytics.nn.modules.conv.Conv [512, 256, 1, 1, None, 1, 1, False]
15 [-2, -1] 1 0 ultralytics.nn.modules.conv.Concat [1]
16 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
17 -1 1 590336 ultralytics.nn.modules.conv.Conv [256, 256, 3, 2]
18 [-1, 12] 1 0 ultralytics.nn.modules.conv.Concat [1]
19 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
20 -1 1 590336 ultralytics.nn.modules.conv.Conv [256, 256, 3, 2]
21 [-1, 7] 1 0 ultralytics.nn.modules.conv.Concat [1]
22 -1 3 2232320 ultralytics.nn.modules.block.RepC3 [512, 256, 3]
23 [16, 19, 22] 1 7303907 ultralytics.nn.modules.head.RTDETRDecoder [1, [256, 256, 256]]
rtdetr-ResNetLayer_APConv summary: 753 layers, 49,690,211 parameters, 49,690,211 gradients, 154.6 GFLOPs