一、本文介绍
本文给大家带来的是主干网络RevColV1,翻译过来就是可逆列网络去发表于ICLR2022,其是一种新型的神经网络设计(和以前的网络结构的传播方式不太一样),由多个子网络(列)通过多级可逆连接组成。这种设计允许在前向传播过程中特征解耦,保持总信息无压缩或丢弃。其非常适合数据集庞大的目标检测任务,数据集数量越多其效果性能越好,亲测在包含1000个图片的数据集上其涨点效果就非常明显了,大家可以多动手尝试,其RevColV2的论文同时已经发布如果代码开源我也会第一时间给大家上传。
推荐指数:⭐⭐⭐⭐⭐
涨点效果:⭐⭐⭐⭐⭐
专栏回顾:YOLOv8改进系列专栏——本专栏持续复习各种顶会内容——科研必备
训练结果对比图->
我这里数据集非常小只有不到一千张图片,同时我用的最微量的版本,涨点幅度还算可以,大家可以尝试其中比较高的版本,大家数据集的图片估计也多效果估计会有一个比较大的提升。
目录
一、本文介绍
二、RevColV1的框架原理
2.1 RevColV1的基本原理
2.1.1 可逆连接设计
2.1.2 特征解耦
2.2 RevColV1的表现
三、RevColV1的核心代码
四、手把手教你添加RevColV1机制
修改一
修改二
修改三
修改四
修改五
修改六
修改七
修改八
五、RevColV1的yaml文件
六、成功运行记录
七、本文总结
二、RevColV1的框架原理
官方论文地址: 官方论文地址
官方代码地址: 官方代码地址
2.1 RevColV1的基本原理
RevCol的主要原理和思想是利用可逆连接来设计网络结构,允许信息在网络的不同分支(列)间自由流动而不丢失。这种多列结构在前向传播过程中逐渐解耦特征,并保持全部信息,而不是进行压缩或舍弃。这样的设计提高了网络在图像分类、对象检测和语义分割等计算机视觉任务中的表现,尤其是在参数量大和数据集大时。
RevCol的创新点我将其总结为以下几点:
1. 可逆连接设计:通过多个子网络(列)间的可逆连接,保证信息在前向传播过程中不丢失。
2. 特征解耦:在每个列中,特征逐渐被解耦,保持总信息而非压缩或舍弃。
3. 适用于大型数据集和参数:在大型数据集和高参数预算下表现出色。
4. 跨模型应用:可作为宏架构方式,应用于变换器或其他神经网络,改善计算机视觉和NLP任务的性能。
简单总结:RevCol通过其独特的多列结构和可逆连接设计,使得网络能够在处理信息时保持完整性,提高特征处理的效率。这种架构在数据丰富且复杂的情况下尤为有效,且可灵活应用于不同类型的神经网络模型中。
其中的创新点第四点不用叙述了,网络结构可以应用于我们的YOLOv8就是最好的印证。
这是论文中的图片1,展示了传统单列网络(a)与RevCol(b)的信息传播对比。在图(a)中,信息通过一个接一个的层线性传播,每层处理后传递给下一层直至输出。而在图(b)中,RevCol通过多个并行列(Col 1 到 Col N)处理信息,其中可逆连接(蓝色曲线)允许信息在列间传递,保持低级别和语义级别的信息传播。这种结构有助于整个网络维持更丰富的信息,并且每个列都能从其他列中学习到信息,增强了特征的表达和网络的学习能力(但是这种做法导致模型的参数量非常巨大,而且训练速度缓慢计算量比较大)。
2.1.1 可逆连接设计
在RevCol中的可逆连接设计允许多个子网络(称为列)之间进行信息的双向流动。这意味着在前向传播的过程中,每一列都能接收到前一列的信息,并将自己的处理结果传递给下一列,同时能够保留传递过程中的所有信息。这种设计避免了在传统的深度网络中常见的信息丢失问题,特别是在网络层次较深时。因此,RevCol可以在深层网络中维持丰富的特征表示,从而提高了模型对数据的表示能力和学习效率。
这张图片展示了RevCol网络的不同组成部分和信息流动方式。
- 图 (a) 展示了RevNet中的一个可逆单元,标识了不同时间步长的状态。
- 图 (b) 展示了多级可逆单元,所有输入在不同级别上进行信息交换。
- 图 (c) 提供了整个可逆列网络架构的概览,其中包含了简化的多级可逆单元。
整个设计允许信息在网络的不同层级和列之间自由流动,而不会丢失任何信息,这对于深层网络的学习和特征提取是非常有益的(我觉得这里有点类似于Neck部分允许层级之间相互交流信息)。
2.1.2 特征解耦
特征解耦是指在RevCol网络的每个子网络(列)中,特征通过可逆连接传递,同时独立地进行处理和学习。这样,每个列都能保持输入信息的完整性,而不会像传统的深度网络那样,在层与层之间传递时压缩或丢弃信息。随着信息在列中的前进,特征之间的关联性逐渐减弱(解耦),使得网络能够更细致地捕捉并强调重要的特征,这有助于提高模型在复杂任务上的性能和泛化能力。
这张图展示了RevCol网络的一个级别(Level l)的微观设计,以及特征融合模块(Fusion Block)的设计。在图(a)中,展示了ConvNeXt级别的标准结构,包括下采样块和残差块。图(b)中的RevCol级别包含了融合模块、残差块和可逆操作。这里的特征解耦是通过融合模块实现的,该模块接收相邻级别的特征图 , 作为输入,并将它们融合以生成新的特征表示。这样,不同级别的特征在融合过程中被解耦,每个级别维持其信息而不压缩或舍弃。图(c)详细描述了融合模块的内部结构,它通过上采样和下采样操作处理不同分辨率的特征图,然后将它们线性叠加,形成为ConvNeXt块提供的特征。这种设计让特征在不同分辨率间流动时进行有效融合。
2.2 RevColV1的表现
这张图片展示了伴随着FLOPs的增长TOP1的准确率情况,可以看出RevColV1伴随着FLOPs的增加效果逐渐明显。
三、RevColV1的核心代码
下面的代码是RevColV1的全部代码,其中包含多个版本,但是大家需要注意这个模型训练非常耗时,参数量非常大,但是其特点就是参数量越大效果越好。其使用方式看章节四。
# --------------------------------------------------------
# Reversible Column Networks
# Copyright (c) 2022 Megvii Inc.
# Licensed under The Apache License 2.0 [see LICENSE for details]
# Written by Yuxuan Cai
# --------------------------------------------------------
from typing import Tuple, Any, List
from timm.models.layers import trunc_normal_
import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import DropPathclass UpSampleConvnext(nn.Module):def __init__(self, ratio, inchannel, outchannel):super().__init__()self.ratio = ratioself.channel_reschedule = nn.Sequential(# LayerNorm(inchannel, eps=1e-6, data_format="channels_last"),nn.Linear(inchannel, outchannel),LayerNorm(outchannel, eps=1e-6, data_format="channels_last"))self.upsample = nn.Upsample(scale_factor=2 ** ratio, mode='nearest')def forward(self, x):x = x.permute(0, 2, 3, 1)x = self.channel_reschedule(x)x = x = x.permute(0, 3, 1, 2)return self.upsample(x)class LayerNorm(nn.Module):r""" LayerNorm that supports two data formats: channels_last (default) or channels_first.The ordering of the dimensions in the inputs. channels_last corresponds to inputs withshape (batch_size, height, width, channels) while channels_first corresponds to inputswith shape (batch_size, channels, height, width)."""def __init__(self, normalized_shape, eps=1e-6, data_format="channels_first", elementwise_affine=True):super().__init__()self.elementwise_affine = elementwise_affineif elementwise_affine:self.weight = nn.Parameter(torch.ones(normalized_shape))self.bias = nn.Parameter(torch.zeros(normalized_shape))self.eps = epsself.data_format = data_formatif self.data_format not in ["channels_last", "channels_first"]:raise NotImplementedErrorself.normalized_shape = (normalized_shape,)def forward(self, x):if self.data_format == "channels_last":return F.layer_norm(x, self.normalized_shape, self.weight, self.bias, self.eps)elif self.data_format == "channels_first":u = x.mean(1, keepdim=True)s = (x - u).pow(2).mean(1, keepdim=True)x = (x - u) / torch.sqrt(s + self.eps)if self.elementwise_affine:x = self.weight[:, None, None] * x + self.bias[:, None, None]return xclass ConvNextBlock(nn.Module):r""" ConvNeXt Block. There are two equivalent implementations:(1) DwConv -> LayerNorm (channels_first) -> 1x1 Conv -> GELU -> 1x1 Conv; all in (N, C, H, W)(2) DwConv -> Permute to (N, H, W, C); LayerNorm (channels_last) -> Linear -> GELU -> Linear; Permute backWe use (2) as we find it slightly faster in PyTorchArgs:dim (int): Number of input channels.drop_path (float): Stochastic depth rate. Default: 0.0layer_scale_init_value (float): Init value for Layer Scale. Default: 1e-6."""def __init__(self, in_channel, hidden_dim, out_channel, kernel_size=3, layer_scale_init_value=1e-6, drop_path=0.0):super().__init__()self.dwconv = nn.Conv2d(in_channel, in_channel, kernel_size=kernel_size, padding=(kernel_size - 1) // 2,groups=in_channel) # depthwise convself.norm = nn.LayerNorm(in_channel, eps=1e-6)self.pwconv1 = nn.Linear(in_channel, hidden_dim) # pointwise/1x1 convs, implemented with linear layersself.act = nn.GELU()self.pwconv2 = nn.Linear(hidden_dim, out_channel)self.gamma = nn.Parameter(layer_scale_init_value * torch.ones((out_channel)),requires_grad=True) if layer_scale_init_value > 0 else Noneself.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()def forward(self, x):input = xx = self.dwconv(x)x = x.permute(0, 2, 3, 1) # (N, C, H, W) -> (N, H, W, C)x = self.norm(x)x = self.pwconv1(x)x = self.act(x)# print(f"x min: {x.min()}, x max: {x.max()}, input min: {input.min()}, input max: {input.max()}, x mean: {x.mean()}, x var: {x.var()}, ratio: {torch.sum(x>8)/x.numel()}")x = self.pwconv2(x)if self.gamma is not None:x = self.gamma * xx = x.permute(0, 3, 1, 2) # (N, H, W, C) -> (N, C, H, W)x = input + self.drop_path(x)return xclass Decoder(nn.Module):def __init__(self, depth=[2, 2, 2, 2], dim=[112, 72, 40, 24], block_type=None, kernel_size=3) -> None:super().__init__()self.depth = depthself.dim = dimself.block_type = block_typeself._build_decode_layer(dim, depth, kernel_size)self.projback = nn.Sequential(nn.Conv2d(in_channels=dim[-1],out_channels=4 ** 2 * 3, kernel_size=1),nn.PixelShuffle(4),)def _build_decode_layer(self, dim, depth, kernel_size):normal_layers = nn.ModuleList()upsample_layers = nn.ModuleList()proj_layers = nn.ModuleList()norm_layer = LayerNormfor i in range(1, len(dim)):module = [self.block_type(dim[i], dim[i], dim[i], kernel_size) for _ in range(depth[i])]normal_layers.append(nn.Sequential(*module))upsample_layers.append(nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True))proj_layers.append(nn.Sequential(nn.Conv2d(dim[i - 1], dim[i], 1, 1),norm_layer(dim[i]),nn.GELU()))self.normal_layers = normal_layersself.upsample_layers = upsample_layersself.proj_layers = proj_layersdef _forward_stage(self, stage, x):x = self.proj_layers[stage](x)x = self.upsample_layers[stage](x)return self.normal_layers[stage](x)def forward(self, c3):x = self._forward_stage(0, c3) # 14x = self._forward_stage(1, x) # 28x = self._forward_stage(2, x) # 56x = self.projback(x)return xclass SimDecoder(nn.Module):def __init__(self, in_channel, encoder_stride) -> None:super().__init__()self.projback = nn.Sequential(LayerNorm(in_channel),nn.Conv2d(in_channels=in_channel,out_channels=encoder_stride ** 2 * 3, kernel_size=1),nn.PixelShuffle(encoder_stride),)def forward(self, c3):return self.projback(c3)def get_gpu_states(fwd_gpu_devices) -> Tuple[List[int], List[torch.Tensor]]:# This will not error out if "arg" is a CPU tensor or a non-tensor type because# the conditionals short-circuit.fwd_gpu_states = []for device in fwd_gpu_devices:with torch.cuda.device(device):fwd_gpu_states.append(torch.cuda.get_rng_state())return fwd_gpu_statesdef get_gpu_device(*args):fwd_gpu_devices = list(set(arg.get_device() for arg in argsif isinstance(arg, torch.Tensor) and arg.is_cuda))return fwd_gpu_devicesdef set_device_states(fwd_cpu_state, devices, states) -> None:torch.set_rng_state(fwd_cpu_state)for device, state in zip(devices, states):with torch.cuda.device(device):torch.cuda.set_rng_state(state)def detach_and_grad(inputs: Tuple[Any, ...]) -> Tuple[torch.Tensor, ...]:if isinstance(inputs, tuple):out = []for inp in inputs:if not isinstance(inp, torch.Tensor):out.append(inp)continuex = inp.detach()x.requires_grad = Trueout.append(x)return tuple(out)else:raise RuntimeError("Only tuple of tensors is supported. Got Unsupported input type: ", type(inputs).__name__)def get_cpu_and_gpu_states(gpu_devices):return torch.get_rng_state(), get_gpu_states(gpu_devices)class ReverseFunction(torch.autograd.Function):@staticmethoddef forward(ctx, run_functions, alpha, *args):l0, l1, l2, l3 = run_functionsalpha0, alpha1, alpha2, alpha3 = alphactx.run_functions = run_functionsctx.alpha = alphactx.preserve_rng_state = Truectx.gpu_autocast_kwargs = {"enabled": torch.is_autocast_enabled(),"dtype": torch.get_autocast_gpu_dtype(),"cache_enabled": torch.is_autocast_cache_enabled()}ctx.cpu_autocast_kwargs = {"enabled": torch.is_autocast_cpu_enabled(),"dtype": torch.get_autocast_cpu_dtype(),"cache_enabled": torch.is_autocast_cache_enabled()}assert len(args) == 5[x, c0, c1, c2, c3] = argsif type(c0) == int:ctx.first_col = Trueelse:ctx.first_col = Falsewith torch.no_grad():gpu_devices = get_gpu_device(*args)ctx.gpu_devices = gpu_devicesctx.cpu_states_0, ctx.gpu_states_0 = get_cpu_and_gpu_states(gpu_devices)c0 = l0(x, c1) + c0 * alpha0ctx.cpu_states_1, ctx.gpu_states_1 = get_cpu_and_gpu_states(gpu_devices)c1 = l1(c0, c2) + c1 * alpha1ctx.cpu_states_2, ctx.gpu_states_2 = get_cpu_and_gpu_states(gpu_devices)c2 = l2(c1, c3) + c2 * alpha2ctx.cpu_states_3, ctx.gpu_states_3 = get_cpu_and_gpu_states(gpu_devices)c3 = l3(c2, None) + c3 * alpha3ctx.save_for_backward(x, c0, c1, c2, c3)return x, c0, c1, c2, c3@staticmethoddef backward(ctx, *grad_outputs):x, c0, c1, c2, c3 = ctx.saved_tensorsl0, l1, l2, l3 = ctx.run_functionsalpha0, alpha1, alpha2, alpha3 = ctx.alphagx_right, g0_right, g1_right, g2_right, g3_right = grad_outputs(x, c0, c1, c2, c3) = detach_and_grad((x, c0, c1, c2, c3))with torch.enable_grad(), \torch.random.fork_rng(devices=ctx.gpu_devices, enabled=ctx.preserve_rng_state), \torch.cuda.amp.autocast(**ctx.gpu_autocast_kwargs), \torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):g3_up = g3_rightg3_left = g3_up * alpha3 ##shortcutset_device_states(ctx.cpu_states_3, ctx.gpu_devices, ctx.gpu_states_3)oup3 = l3(c2, None)torch.autograd.backward(oup3, g3_up, retain_graph=True)with torch.no_grad():c3_left = (1 / alpha3) * (c3 - oup3) ## feature reverseg2_up = g2_right + c2.gradg2_left = g2_up * alpha2 ##shortcut(c3_left,) = detach_and_grad((c3_left,))set_device_states(ctx.cpu_states_2, ctx.gpu_devices, ctx.gpu_states_2)oup2 = l2(c1, c3_left)torch.autograd.backward(oup2, g2_up, retain_graph=True)c3_left.requires_grad = Falsecout3 = c3_left * alpha3 ##alpha3 updatetorch.autograd.backward(cout3, g3_up)with torch.no_grad():c2_left = (1 / alpha2) * (c2 - oup2) ## feature reverseg3_left = g3_left + c3_left.grad if c3_left.grad is not None else g3_leftg1_up = g1_right + c1.gradg1_left = g1_up * alpha1 ##shortcut(c2_left,) = detach_and_grad((c2_left,))set_device_states(ctx.cpu_states_1, ctx.gpu_devices, ctx.gpu_states_1)oup1 = l1(c0, c2_left)torch.autograd.backward(oup1, g1_up, retain_graph=True)c2_left.requires_grad = Falsecout2 = c2_left * alpha2 ##alpha2 updatetorch.autograd.backward(cout2, g2_up)with torch.no_grad():c1_left = (1 / alpha1) * (c1 - oup1) ## feature reverseg0_up = g0_right + c0.gradg0_left = g0_up * alpha0 ##shortcutg2_left = g2_left + c2_left.grad if c2_left.grad is not None else g2_left ## Fusion(c1_left,) = detach_and_grad((c1_left,))set_device_states(ctx.cpu_states_0, ctx.gpu_devices, ctx.gpu_states_0)oup0 = l0(x, c1_left)torch.autograd.backward(oup0, g0_up, retain_graph=True)c1_left.requires_grad = Falsecout1 = c1_left * alpha1 ##alpha1 updatetorch.autograd.backward(cout1, g1_up)with torch.no_grad():c0_left = (1 / alpha0) * (c0 - oup0) ## feature reversegx_up = x.grad ## Fusiong1_left = g1_left + c1_left.grad if c1_left.grad is not None else g1_left ## Fusionc0_left.requires_grad = Falsecout0 = c0_left * alpha0 ##alpha0 updatetorch.autograd.backward(cout0, g0_up)if ctx.first_col:return None, None, gx_up, None, None, None, Noneelse:return None, None, gx_up, g0_left, g1_left, g2_left, g3_leftclass Fusion(nn.Module):def __init__(self, level, channels, first_col) -> None:super().__init__()self.level = levelself.first_col = first_colself.down = nn.Sequential(nn.Conv2d(channels[level - 1], channels[level], kernel_size=2, stride=2),LayerNorm(channels[level], eps=1e-6, data_format="channels_first"),) if level in [1, 2, 3] else nn.Identity()if not first_col:self.up = UpSampleConvnext(1, channels[level + 1], channels[level]) if level in [0, 1, 2] else nn.Identity()def forward(self, *args):c_down, c_up = argsif self.first_col:x = self.down(c_down)return xif self.level == 3:x = self.down(c_down)else:x = self.up(c_up) + self.down(c_down)return xclass Level(nn.Module):def __init__(self, level, channels, layers, kernel_size, first_col, dp_rate=0.0) -> None:super().__init__()countlayer = sum(layers[:level])expansion = 4self.fusion = Fusion(level, channels, first_col)modules = [ConvNextBlock(channels[level], expansion * channels[level], channels[level], kernel_size=kernel_size,layer_scale_init_value=1e-6, drop_path=dp_rate[countlayer + i]) for i inrange(layers[level])]self.blocks = nn.Sequential(*modules)def forward(self, *args):x = self.fusion(*args)x = self.blocks(x)return xclass SubNet(nn.Module):def __init__(self, channels, layers, kernel_size, first_col, dp_rates, save_memory) -> None:super().__init__()shortcut_scale_init_value = 0.5self.save_memory = save_memoryself.alpha0 = nn.Parameter(shortcut_scale_init_value * torch.ones((1, channels[0], 1, 1)),requires_grad=True) if shortcut_scale_init_value > 0 else Noneself.alpha1 = nn.Parameter(shortcut_scale_init_value * torch.ones((1, channels[1], 1, 1)),requires_grad=True) if shortcut_scale_init_value > 0 else Noneself.alpha2 = nn.Parameter(shortcut_scale_init_value * torch.ones((1, channels[2], 1, 1)),requires_grad=True) if shortcut_scale_init_value > 0 else Noneself.alpha3 = nn.Parameter(shortcut_scale_init_value * torch.ones((1, channels[3], 1, 1)),requires_grad=True) if shortcut_scale_init_value > 0 else Noneself.level0 = Level(0, channels, layers, kernel_size, first_col, dp_rates)self.level1 = Level(1, channels, layers, kernel_size, first_col, dp_rates)self.level2 = Level(2, channels, layers, kernel_size, first_col, dp_rates)self.level3 = Level(3, channels, layers, kernel_size, first_col, dp_rates)def _forward_nonreverse(self, *args):x, c0, c1, c2, c3 = argsc0 = (self.alpha0) * c0 + self.level0(x, c1)c1 = (self.alpha1) * c1 + self.level1(c0, c2)c2 = (self.alpha2) * c2 + self.level2(c1, c3)c3 = (self.alpha3) * c3 + self.level3(c2, None)return c0, c1, c2, c3def _forward_reverse(self, *args):local_funs = [self.level0, self.level1, self.level2, self.level3]alpha = [self.alpha0, self.alpha1, self.alpha2, self.alpha3]_, c0, c1, c2, c3 = ReverseFunction.apply(local_funs, alpha, *args)return c0, c1, c2, c3def forward(self, *args):self._clamp_abs(self.alpha0.data, 1e-3)self._clamp_abs(self.alpha1.data, 1e-3)self._clamp_abs(self.alpha2.data, 1e-3)self._clamp_abs(self.alpha3.data, 1e-3)if self.save_memory:return self._forward_reverse(*args)else:return self._forward_nonreverse(*args)def _clamp_abs(self, data, value):with torch.no_grad():sign = data.sign()data.abs_().clamp_(value)data *= signclass Classifier(nn.Module):def __init__(self, in_channels, num_classes):super().__init__()self.avgpool = nn.AdaptiveAvgPool2d((1, 1))self.classifier = nn.Sequential(nn.LayerNorm(in_channels, eps=1e-6), # final norm layernn.Linear(in_channels, num_classes),)def forward(self, x):x = self.avgpool(x)x = x.view(x.size(0), -1)x = self.classifier(x)return xclass FullNet(nn.Module):def __init__(self, channels=[32, 64, 96, 128], layers=[2, 3, 6, 3], num_subnet=5, kernel_size=3, drop_path=0.0,save_memory=True, inter_supv=True) -> None:super().__init__()self.num_subnet = num_subnetself.inter_supv = inter_supvself.channels = channelsself.layers = layersself.stem = nn.Sequential(nn.Conv2d(3, channels[0], kernel_size=4, stride=4),LayerNorm(channels[0], eps=1e-6, data_format="channels_first"))dp_rate = [x.item() for x in torch.linspace(0, drop_path, sum(layers))]for i in range(num_subnet):first_col = True if i == 0 else Falseself.add_module(f'subnet{str(i)}', SubNet(channels, layers, kernel_size, first_col, dp_rates=dp_rate, save_memory=save_memory))self.apply(self._init_weights)self.width_list = [i.size(1) for i in self.forward(torch.randn(1, 3, 640, 640))]def forward(self, x):c0, c1, c2, c3 = 0, 0, 0, 0x = self.stem(x)for i in range(self.num_subnet):c0, c1, c2, c3 = getattr(self, f'subnet{str(i)}')(x, c0, c1, c2, c3)return [c0, c1, c2, c3]def _init_weights(self, module):if isinstance(module, nn.Conv2d):trunc_normal_(module.weight, std=.02)nn.init.constant_(module.bias, 0)elif isinstance(module, nn.Linear):trunc_normal_(module.weight, std=.02)nn.init.constant_(module.bias, 0)##-------------------------------------- Tiny -----------------------------------------def revcol_tiny(save_memory=True, inter_supv=True, drop_path=0.1, kernel_size=3):channels = [64, 128, 256, 512]layers = [2, 2, 4, 2]num_subnet = 4return FullNet(channels, layers, num_subnet, drop_path=drop_path, save_memory=save_memory, inter_supv=inter_supv,kernel_size=kernel_size)##-------------------------------------- Small -----------------------------------------def revcol_small(save_memory=True, inter_supv=True, drop_path=0.3, kernel_size=3):channels = [64, 128, 256, 512]layers = [2, 2, 4, 2]num_subnet = 8return FullNet(channels, layers, num_subnet, drop_path=drop_path, save_memory=save_memory, inter_supv=inter_supv,kernel_size=kernel_size)##-------------------------------------- Base -----------------------------------------def revcol_base(save_memory=True, inter_supv=True, drop_path=0.4, kernel_size=3, head_init_scale=None):channels = [72, 144, 288, 576]layers = [1, 1, 3, 2]num_subnet = 16return FullNet(channels, layers, num_subnet, drop_path=drop_path, save_memory=save_memory, inter_supv=inter_supv, kernel_size=kernel_size)##-------------------------------------- Large -----------------------------------------def revcol_large(save_memory=True, inter_supv=True, drop_path=0.5, kernel_size=3, head_init_scale=None):channels = [128, 256, 512, 1024]layers = [1, 2, 6, 2]num_subnet = 8return FullNet(channels, layers, num_subnet, drop_path=drop_path, save_memory=save_memory, inter_supv=inter_supv, kernel_size=kernel_size)##--------------------------------------Extra-Large -----------------------------------------
def revcol_xlarge(save_memory=True, inter_supv=True, drop_path=0.5, kernel_size=3, head_init_scale=None):channels = [224, 448, 896, 1792]layers = [1, 2, 6, 2]num_subnet = 8return FullNet(channels, layers, num_subnet, drop_path=drop_path, save_memory=save_memory, inter_supv=inter_supv, kernel_size=kernel_size)# model = revcol_xlarge(True)
# # 示例输入
# input = torch.randn(64, 3, 224, 224)
# output = model(input)
#
# print(len(output))#torch.Size([3, 64, 224, 224])
四、手把手教你添加RevColV1机制
这个主干的网络结构添加起来算是所有的改进机制里最麻烦的了,因为有一些网略结构可以用yaml文件搭建出来,有一些网络结构其中的一些细节根本没有办法用yaml文件去搭建,用yaml文件去搭建会损失一些细节部分(而且一个网络结构设计很多细节的结构修改方式都不一样,一个一个去修改大家难免会出错),所以这里让网络直接返回整个网络,然后修改部分 yolo代码以后就都以这种形式添加了,以后我提出的网络模型基本上都会通过这种方式修改,我也会进行一些模型细节改进。创新出新的网络结构大家直接拿来用就可以的。下面开始添加教程->
(同时每一个后面都有代码,大家拿来复制粘贴替换即可,但是要看好了不要复制粘贴替换多了)
修改一
我们复制网络结构代码到“ultralytics/nn/modules”目录下创建一个py文件复制粘贴进去 ,我这里起的名字是RevColV1。
修改二
找到如下的文件"ultralytics/nn/tasks.py" 在开始的部分导入我们的模型如下图。
from .modules.RevColV1 import revcol_tiny, revcol_base, revcol_xlarge, revcol_large, revcol_small
修改三
添加如下两行代码!!!
修改四
找到七百多行大概把具体看图片,按照图片来修改就行,添加红框内的部分,注意没有()只是函数名,我这里只添加了部分的版本,大家有兴趣这个RevColV1还有更多的版本可以添加,看我给的代码函数头即可。
elif m in {自行添加对应的模型即可,下面都是一样的}:m = m()c2 = m.width_list # 返回通道列表backbone = True
修改五
下面的两个红框内都是需要改动的。
if isinstance(c2, list):m_ = mm_.backbone = Trueelse:m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args) # modulet = str(m)[8:-2].replace('__main__.', '') # module typem.np = sum(x.numel() for x in m_.parameters()) # number paramsm_.i, m_.f, m_.type = i + 4 if backbone else i, f, t # attach index, 'from' index, type
修改六
如下的也需要修改,全部按照我的来。
代码如下把原先的代码替换了即可。
if verbose:LOGGER.info(f'{i:>3}{str(f):>20}{n_:>3}{m.np:10.0f} {t:<45}{str(args):<30}') # printsave.extend(x % (i + 4 if backbone else i) for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelistlayers.append(m_)if i == 0:ch = []if isinstance(c2, list):ch.extend(c2)if len(c2) != 5:ch.insert(0, 0)else:ch.append(c2)
修改七
修改七和前面的都不太一样,需要修改前向传播中的一个部分, 已经离开了parse_model方法了。
可以在图片中开代码行数,没有离开task.py文件都是同一个文件。 同时这个部分有好几个前向传播都很相似,大家不要看错了,是70多行左右的!!!,同时我后面提供了代码,大家直接复制粘贴即可,有时间我针对这里会出一个视频。
代码如下->
def _predict_once(self, x, profile=False, visualize=False):"""Perform a forward pass through the network.Args:x (torch.Tensor): The input tensor to the model.profile (bool): Print the computation time of each layer if True, defaults to False.visualize (bool): Save the feature maps of the model if True, defaults to False.Returns:(torch.Tensor): The last output of the model."""y, dt = [], [] # outputsfor m in self.model:if m.f != -1: # if not from previous layerx = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layersif profile:self._profile_one_layer(m, x, dt)if hasattr(m, 'backbone'):x = m(x)if len(x) != 5: # 0 - 5x.insert(0, None)for index, i in enumerate(x):if index in self.save:y.append(i)else:y.append(None)x = x[-1] # 最后一个输出传给下一层else:x = m(x) # runy.append(x if m.i in self.save else None) # save outputif visualize:feature_visualization(x, m.type, m.i, save_dir=visualize)return x
到这里就完成了修改部分,但是这里面细节很多,大家千万要注意不要替换多余的代码,导致报错,也不要拉下任何一部,都会导致运行失败,而且报错很难排查!!!很难排查!!!
修改八
我们找到如下文件'ultralytics/utils/torch_utils.py'按照如下的图片进行修改,否则容易打印不出来计算量。
五、RevColV1的yaml文件
复制如下yaml文件进行运行!!!
# 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 GFLOPss: [0.33, 0.50, 1024] # YOLOv8s summary: 225 layers, 11166560 parameters, 11166544 gradients, 28.8 GFLOPsm: [0.67, 0.75, 768] # YOLOv8m summary: 295 layers, 25902640 parameters, 25902624 gradients, 79.3 GFLOPsl: [1.00, 1.00, 512] # YOLOv8l summary: 365 layers, 43691520 parameters, 43691504 gradients, 165.7 GFLOPsx: [1.00, 1.25, 512] # YOLOv8x summary: 365 layers, 68229648 parameters, 68229632 gradients, 258.5 GFLOP# YOLOv8.0n backbone
backbone:# [from, repeats, module, args]- [-1, 1, revcol_tiny, []] # 4- [-1, 1, SPPF, [1024, 5]] # 5# YOLOv8.0n head
head:- [-1, 1, nn.Upsample, [None, 2, 'nearest']] # 6- [[-1, 3], 1, Concat, [1]] # 7 cat backbone P4- [-1, 3, C2f, [512]] # 8- [-1, 1, nn.Upsample, [None, 2, 'nearest']] # 9- [[-1, 2], 1, Concat, [1]] # 10 cat backbone P3- [-1, 3, C2f, [256]] # 11 (P3/8-small)- [-1, 1, Conv, [256, 3, 2]] # 12- [[-1, 8], 1, Concat, [1]] # 13 cat head P4- [-1, 3, C2f, [512]] # 14 (P4/16-medium)- [-1, 1, Conv, [512, 3, 2]] # 15- [[-1, 5], 1, Concat, [1]] # 16 cat head P5- [-1, 3, C2f, [1024]] # 17 (P5/32-large)- [[11, 14, 17], 1, Detect, [nc]] # Detect(P3, P4, P5)
六、成功运行记录
下面是成功运行的截图,已经完成了有1个epochs的训练,图片太大截不全第2个epochs了。
七、本文总结
到此本文的正式分享内容就结束了,在这里给大家推荐我的YOLOv8改进有效涨点专栏,本专栏目前为新开的平均质量分98分,后期我会根据各种最新的前沿顶会进行论文复现,也会对一些老的改进机制进行补充,目前本专栏免费阅读(暂时,大家尽早关注不迷路~),如果大家觉得本文帮助到你了,订阅本专栏,关注后续更多的更新~
专栏回顾:YOLOv8改进系列专栏——本专栏持续复习各种顶会内容——科研必备