目标检测算法改进系列之Backbone替换为FocalNet

最新推荐文章于 2024-05-03 10:14:52 发布

我悟了-

最新推荐文章于 2024-05-03 10:14:52 发布

阅读量432

点赞数 2

分类专栏：目标检测算法改进系列文章标签：目标检测算法人工智能计算机视觉 python pytorch YOLO

本文链接：https://blog.csdn.net/dm_zx/article/details/133587650

版权

目标检测算法改进系列专栏收录该内容

23 篇文章 7 订阅

订阅专栏

FocalNet

近些年，Transformers在自然语言处理、图像分类、目标检测和图像分割上均取得了较大的成功，归根结底是自注意力（SA ：self-attention）起到了关键性的作用，因此能够支持输入信息的全局交互。但是由于视觉tokens的大量存在，自注意力的计算复杂度高，尤其是在高分辨的输入时，因此针对该缺陷，论文《Focal Modulation Networks》提出了FocalNet网络。

论文地址：Focal Modulation Networks

原理：使用新提出的Focal Modulation替代之前的SA自注意力模块，解耦聚合和单个查询过程，先将查询周围的上下文信息进行聚合，再根据聚合信息获取查询结果。如下图所示，图中红色表示query token。对比来看，Window-wise Self-Attention (SA)利用周围的token（橙色）来捕获空间上下文信息；在此基础上，Focal Attention扩大了感受野，还可以使用更远的summarized tokens（蓝色）；而Focal Modulation更为强大，先利用诸如depth-wise convolution的方式将不同粒度级别的空间上下文编码为summarized tokens （橙色、绿色和蓝色），再根据查询内容，选择性的将这些summarized tokens融合为query token。而本文新提出的方式便是进行轻量化，将聚合和单个查询进行解耦，减少计算量。

在前两者中，绿色和紫色箭头分别代表注意力交互和基于查询的聚合，但是都存在一个缺陷，即：均需要涉及大量的交互和聚合操作。而Focal Modulation计算过程得到大量简化。
原理图

FocalNet代码实现

# --------------------------------------------------------
# FocalNets -- Focal Modulation Networks
# Copyright (c) 2022 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Jianwei Yang (jianwyan@microsoft.com)
# --------------------------------------------------------

import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
from timm.models.layers import DropPath, to_2tuple, trunc_normal_

__all__ = ['focalnet_tiny_srf', 'focalnet_tiny_lrf', 'focalnet_small_srf', 'focalnet_small_lrf', 'focalnet_base_srf', 'focalnet_base_lrf', 'focalnet_large_fl3', 'focalnet_large_fl4', 'focalnet_xlarge_fl3', 'focalnet_xlarge_fl4', 'focalnet_huge_fl3', 'focalnet_huge_fl4']

def update_weight(model_dict, weight_dict):
    idx, temp_dict = 0, {}
    for k, v in weight_dict.items():
        if k in model_dict.keys() and np.shape(model_dict[k]) == np.shape(v):
            temp_dict[k] = v
            idx += 1
    model_dict.update(temp_dict)
    print(f'loading weights... {idx}/{len(model_dict)} items')
    return model_dict

class Mlp(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop)

    def forward(self, x):
        x = self.fc1(x)     
        x = self.act(x)
        x = self.drop(x)
        x = self.fc2(x)
        x = self.drop(x)
        return x

class FocalModulation(nn.Module):
    def __init__(self, dim, focal_window, focal_level, focal_factor=2, bias=True, proj_drop=0., use_postln_in_modulation=False, normalize_modulator=False):
        super().__init__()

        self.dim = dim
        self.focal_window = focal_window
        self.focal_level = focal_level
        self.focal_factor = focal_factor
        self.use_postln_in_modulation = use_postln_in_modulation
        self.normalize_modulator = normalize_modulator

        self.f = nn.Linear(dim, 2*dim + (self.focal_level+1), bias=bias)
        self.h = nn.Conv2d(dim, dim, kernel_size=1, stride=1, bias=bias)

        self.act = nn.GELU()
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)
        self.focal_layers = nn.ModuleList()
                
        self.kernel_sizes = []
        for k in range(self.focal_level):
            kernel_size = self.focal_factor*k + self.focal_window
            self.focal_layers.append(
                nn.Sequential(
                    nn.Conv2d(dim, dim, kernel_size=kernel_size, stride=1, 
                    groups=dim, padding=kernel_size//2, bias=False),
                    nn.GELU(),
                    )
                )              
            self.kernel_sizes.append(kernel_size)          
        if self.use_postln_in_modulation:
            self.ln = nn.LayerNorm(dim)

    def forward(self, x):
        """
        Args:
            x: input features with shape of (B, H, W, C)
        """
        C = x.shape[-1]

        # pre linear projection
        x = self.f(x).permute(0, 3, 1, 2).contiguous()
        q, ctx, gates = torch.split(x, (C, C, self.focal_level+1), 1)
        
        # context aggreation
        ctx_all = 0 
        for l in range(self.focal_level):         
            ctx = self.focal_layers[l](ctx)
            ctx_all = ctx_all + ctx * gates[:, l:l+1]
        ctx_global = self.act(ctx.mean(2, keepdim=True).mean(3, keepdim=True))
        ctx_all = ctx_all + ctx_global * gates[:,self.focal_level:]

        # normalize context
        if self.normalize_modulator:
            ctx_all = ctx_all / (self.focal_level+1)

        # focal modulation
        modulator = self.h(ctx_all)
        x_out = q * modulator
        x_out = x_out.permute(0, 2, 3, 1).contiguous()
        if self.use_postln_in_modulation:
            x_out = self.ln(x_out)
        
        # post linear porjection
        x_out = self.proj(x_out)
        x_out = self.proj_drop(x_out)
        return x_out

    def extra_repr(self) -> str:
        return f'dim={self.dim}'

    def flops(self, N):
        # calculate flops for 1 window with token length of N
        flops = 0

        flops += N * self.dim * (self.dim * 2 + (self.focal_level+1))

        # focal convolution
        for k in range(self.focal_level):
            flops += N * (self.kernel_sizes[k]**2+1) * self.dim

        # global gating
        flops += N * 1 * self.dim 

        #  self.linear
        flops += N * self.dim * (self.dim + 1)

        # x = self.proj(x)
        flops += N * self.dim * self.dim
        return flops

class FocalNetBlock(nn.Module):
    r""" Focal Modulation Network Block.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resulotion.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        drop (float, optional): Dropout rate. Default: 0.0
        drop_path (float, optional): Stochastic depth rate. Default: 0.0
        act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
        focal_level (int): Number of focal levels. 
        focal_window (int): Focal window size at first focal level
        use_layerscale (bool): Whether use layerscale
        layerscale_value (float): Initial layerscale value
        use_postln (bool): Whether use layernorm after modulation
    """

    def __init__(self, dim, input_resolution, mlp_ratio=4., drop=0., drop_path=0., 
                    act_layer=nn.GELU, norm_layer=nn.LayerNorm,
                    focal_level=1, focal_window=3,
                    use_layerscale=False, layerscale_value=1e-4, 
                    use_postln=False, use_postln_in_modulation=False, 
                    normalize_modulator=False):
        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.mlp_ratio = mlp_ratio

        self.focal_window = focal_window
        self.focal_level = focal_level
        self.use_postln = use_postln

        self.norm1 = norm_layer(dim)
        self.modulation = FocalModulation(
            dim, proj_drop=drop, focal_window=focal_window, focal_level=self.focal_level, 
            use_postln_in_modulation=use_postln_in_modulation, normalize_modulator=normalize_modulator
        )

        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

        self.gamma_1 = 1.0
        self.gamma_2 = 1.0    
        if use_layerscale:
            self.gamma_1 = nn.Parameter(layerscale_value * torch.ones((dim)), requires_grad=True)
            self.gamma_2 = nn.Parameter(layerscale_value * torch.ones((dim)), requires_grad=True)

        self.H = None
        self.W = None

    def forward(self, x):
        H, W = self.H, self.W
        B, L, C = x.shape
        shortcut = x

        # Focal Modulation
        x = x if self.use_postln else self.norm1(x)
        x = x.view(B, H, W, C)
        x = self.modulation(x).view(B, H * W, C)
        x = x if not self.use_postln else self.norm1(x)

        # FFN
        x = shortcut + self.drop_path(self.gamma_1 * x)
        x = x + self.drop_path(self.gamma_2 * (self.norm2(self.mlp(x)) if self.use_postln else self.mlp(self.norm2(x))))

        return x

    def extra_repr(self) -> str:
        return f"dim={self.dim}, input_resolution={self.input_resolution}, " \
               f"mlp_ratio={self.mlp_ratio}"

    def flops(self):
        flops = 0
        H, W = self.input_resolution
        # norm1
        flops += self.dim * H * W
        
        # W-MSA/SW-MSA
        flops += self.modulation.flops(H*W)

        # mlp
        flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio
        # norm2
        flops += self.dim * H * W
        return flops

class BasicLayer(nn.Module):
    """ A basic Focal Transformer layer for one stage.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resolution.
        depth (int): Number of blocks.
        window_size (int): Local window size.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
        norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
        focal_level (int): Number of focal levels
        focal_window (int): Focal window size at first focal level
        use_layerscale (bool): Whether use layerscale
        layerscale_value (float): Initial layerscale value
        use_postln (bool): Whether use layernorm after modulation
    """

    def __init__(self, dim, out_dim, input_resolution, depth,
                 mlp_ratio=4., drop=0., drop_path=0., norm_layer=nn.LayerNorm, 
                 downsample=None, use_checkpoint=False, 
                 focal_level=1, focal_window=1, 
                 use_conv_embed=False, 
                 use_layerscale=False, layerscale_value=1e-4, 
                 use_postln=False, 
                 use_postln_in_modulation=False, 
                 normalize_modulator=False):

        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.depth = depth
        self.use_checkpoint = use_checkpoint
        
        # build blocks
        self.blocks = nn.ModuleList([
            FocalNetBlock(
                dim=dim, 
                input_resolution=input_resolution,
                mlp_ratio=mlp_ratio, 
                drop=drop, 
                drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
                norm_layer=norm_layer,
                focal_level=focal_level,
                focal_window=focal_window, 
                use_layerscale=use_layerscale, 
                layerscale_value=layerscale_value,
                use_postln=use_postln, 
                use_postln_in_modulation=use_postln_in_modulation, 
                normalize_modulator=normalize_modulator, 
            )
            for i in range(depth)])

        if downsample is not None:
            self.downsample = downsample(
                img_size=input_resolution, 
                patch_size=2, 
                in_chans=dim, 
                embed_dim=out_dim, 
                use_conv_embed=use_conv_embed, 
                norm_layer=norm_layer, 
                is_stem=False
            )
        else:
            self.downsample = None

    def forward(self, x, H, W):
        for blk in self.blocks:
            blk.H, blk.W = H, W
            if self.use_checkpoint:
                x = checkpoint.checkpoint(blk, x)
            else:
                x = blk(x)

        if self.downsample is not None:
            x = x.transpose(1, 2).reshape(x.shape[0], -1, H, W)
            x, Ho, Wo = self.downsample(x)
        else:
            Ho, Wo = H, W        
        return x, Ho, Wo

    def extra_repr(self) -> str:
        return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}"

    def flops(self):
        flops = 0
        for blk in self.blocks:
            flops += blk.flops()
        if self.downsample is not None:
            flops += self.downsample.flops()
        return flops

class PatchEmbed(nn.Module):
    r""" Image to Patch Embedding

    Args:
        img_size (int): Image size.  Default: 224.
        patch_size (int): Patch token size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    def __init__(self, img_size=(224, 224), patch_size=4, in_chans=3, embed_dim=96, use_conv_embed=False, norm_layer=None, is_stem=False):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]]
        self.img_size = img_size
        self.patch_size = patch_size
        self.patches_resolution = patches_resolution
        self.num_patches = patches_resolution[0] * patches_resolution[1]

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        if use_conv_embed:
            # if we choose to use conv embedding, then we treat the stem and non-stem differently
            if is_stem:
                kernel_size = 7; padding = 2; stride = 4
            else:
                kernel_size = 3; padding = 1; stride = 2
            self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=kernel_size, stride=stride, padding=padding)
        else:
            self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    def forward(self, x):
        B, C, H, W = x.shape

        x = self.proj(x)        
        H, W = x.shape[2:]
        x = x.flatten(2).transpose(1, 2)  # B Ph*Pw C
        if self.norm is not None:
            x = self.norm(x)
        return x, H, W

    def flops(self):
        Ho, Wo = self.patches_resolution
        flops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1])
        if self.norm is not None:
            flops += Ho * Wo * self.embed_dim
        return flops

class FocalNet(nn.Module):
    r""" Focal Modulation Networks (FocalNets)

    Args:
        img_size (int | tuple(int)): Input image size. Default 224
        patch_size (int | tuple(int)): Patch size. Default: 4
        in_chans (int): Number of input image channels. Default: 3
        num_classes (int): Number of classes for classification head. Default: 1000
        embed_dim (int): Patch embedding dimension. Default: 96
        depths (tuple(int)): Depth of each Focal Transformer layer.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4
        drop_rate (float): Dropout rate. Default: 0
        drop_path_rate (float): Stochastic depth rate. Default: 0.1
        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
        patch_norm (bool): If True, add normalization after patch embedding. Default: True
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False 
        focal_levels (list): How many focal levels at all stages. Note that this excludes the finest-grain level. Default: [1, 1, 1, 1] 
        focal_windows (list): The focal window size at all stages. Default: [7, 5, 3, 1] 
        use_conv_embed (bool): Whether use convolutional embedding. We noted that using convolutional embedding usually improve the performance, but we do not use it by default. Default: False 
        use_layerscale (bool): Whether use layerscale proposed in CaiT. Default: False 
        layerscale_value (float): Value for layer scale. Default: 1e-4 
        use_postln (bool): Whether use layernorm after modulation (it helps stablize training of large models)
    """
    def __init__(self, 
                img_size=224, 
                patch_size=4, 
                in_chans=3, 
                num_classes=1000,
                embed_dim=96, 
                depths=[2, 2, 6, 2], 
                mlp_ratio=4., 
                drop_rate=0., 
                drop_path_rate=0.1,
                norm_layer=nn.LayerNorm, 
                patch_norm=True,
                use_checkpoint=False,                 
                focal_levels=[2, 2, 2, 2], 
                focal_windows=[3, 3, 3, 3], 
                use_conv_embed=False, 
                use_layerscale=False, 
                layerscale_value=1e-4, 
                use_postln=False, 
                use_postln_in_modulation=False, 
                normalize_modulator=False, 
                **kwargs):
        super().__init__()

        self.num_layers = len(depths)
        embed_dim = [embed_dim * (2 ** i) for i in range(self.num_layers)]

        self.num_classes = num_classes
        self.embed_dim = embed_dim
        self.patch_norm = patch_norm
        self.num_features = embed_dim[-1]
        self.mlp_ratio = mlp_ratio
        
        # split image into patches using either non-overlapped embedding or overlapped embedding
        self.patch_embed = PatchEmbed(
            img_size=to_2tuple(img_size), 
            patch_size=patch_size, 
            in_chans=in_chans, 
            embed_dim=embed_dim[0], 
            use_conv_embed=use_conv_embed, 
            norm_layer=norm_layer if self.patch_norm else None, 
            is_stem=True)

        num_patches = self.patch_embed.num_patches
        patches_resolution = self.patch_embed.patches_resolution
        self.patches_resolution = patches_resolution
        self.pos_drop = nn.Dropout(p=drop_rate)

        # stochastic depth
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule

        # build layers
        self.layers = nn.ModuleList()
        for i_layer in range(self.num_layers):
            layer = BasicLayer(dim=embed_dim[i_layer], 
                               out_dim=embed_dim[i_layer+1] if (i_layer < self.num_layers - 1) else None,  
                               input_resolution=(patches_resolution[0] // (2 ** i_layer),
                                                 patches_resolution[1] // (2 ** i_layer)),
                               depth=depths[i_layer],
                               mlp_ratio=self.mlp_ratio,
                               drop=drop_rate, 
                               drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
                               norm_layer=norm_layer, 
                               downsample=PatchEmbed if (i_layer < self.num_layers - 1) else None,
                               focal_level=focal_levels[i_layer], 
                               focal_window=focal_windows[i_layer], 
                               use_conv_embed=use_conv_embed,
                               use_checkpoint=use_checkpoint, 
                               use_layerscale=use_layerscale, 
                               layerscale_value=layerscale_value, 
                               use_postln=use_postln,
                               use_postln_in_modulation=use_postln_in_modulation, 
                               normalize_modulator=normalize_modulator
                    )
            self.layers.append(layer)

        self.norm = norm_layer(self.num_features)

        self.apply(self._init_weights)
        self.channel = [i.size(1) for i in self.forward(torch.randn(1, 3, 640, 640))]

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    @torch.jit.ignore
    def no_weight_decay(self):
        return {''}

    @torch.jit.ignore
    def no_weight_decay_keywords(self):
        return {''}

    def forward(self, x):
        input_size = x.size(2)
        scale = [4, 8, 16, 32]
        
        x, H, W = self.patch_embed(x)
        x = self.pos_drop(x)
        features = [x, None, None, None]
        for layer in self.layers:
            x, H, W = layer(x, H, W)
            if input_size // H in scale:
                features[scale.index(input_size // H)] = x
        # features[-1] = self.norm(features[-1])  # B L C
        
        for i in range(len(features)):
            features[i] = torch.transpose(features[i], dim0=2, dim1=1).view(-1,features[i].size(2), int(features[i].size(1) ** 0.5), int(features[i].size(1) ** 0.5))
        
        return features

    def flops(self):
        flops = 0
        flops += self.patch_embed.flops()
        for i, layer in enumerate(self.layers):
            flops += layer.flops()
        flops += self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // (2 ** self.num_layers)
        flops += self.num_features * self.num_classes
        return flops

model_urls = {
    "focalnet_tiny_srf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_tiny_srf.pth",
    "focalnet_tiny_lrf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_tiny_lrf.pth",
    "focalnet_small_srf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_small_srf.pth",
    "focalnet_small_lrf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_small_lrf.pth",
    "focalnet_base_srf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_base_srf.pth",
    "focalnet_base_lrf": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_base_lrf.pth",    
    "focalnet_large_fl3": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384.pth", 
    "focalnet_large_fl4": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_large_lrf_384_fl4.pth", 
    "focalnet_xlarge_fl3": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_xlarge_lrf_384.pth", 
    "focalnet_xlarge_fl4": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_xlarge_lrf_384_fl4.pth", 
    "focalnet_huge_fl3": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_huge_lrf_224.pth", 
    "focalnet_huge_fl4": "https://projects4jw.blob.core.windows.net/focalnet/release/classification/focalnet_huge_lrf_224_fl4.pth", 
}

def focalnet_tiny_srf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 6, 2], embed_dim=96, **kwargs)
    if pretrained:
        url = model_urls['focalnet_tiny_srf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu", check_hash=True)
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_small_srf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=96, **kwargs)
    if pretrained:
        url = model_urls['focalnet_small_srf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_base_srf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=128, **kwargs)
    if pretrained:
        url = model_urls['focalnet_base_srf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_tiny_lrf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 6, 2], embed_dim=96, **kwargs)
    if pretrained:
        url = model_urls['focalnet_tiny_lrf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu", check_hash=True)
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_small_lrf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=96, **kwargs)
    if pretrained:
        url = model_urls['focalnet_small_lrf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_base_lrf(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=128, **kwargs)
    if pretrained:
        url = model_urls['focalnet_base_lrf']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_tiny_iso(pretrained=False, **kwargs):
    model = FocalNet(depths=[12], patch_size=16, embed_dim=192, **kwargs)
    if pretrained:
        url = model_urls['focalnet_tiny_iso']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu", check_hash=True)
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_small_iso(pretrained=False, **kwargs):
    model = FocalNet(depths=[12], patch_size=16, embed_dim=384, **kwargs)
    if pretrained:
        url = model_urls['focalnet_small_iso']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_base_iso(pretrained=False, **kwargs):
    model = FocalNet(depths=[12], patch_size=16, embed_dim=768, focal_levels=[3], focal_windows=[3], use_layerscale=True, use_postln=True, **kwargs)
    if pretrained:
        url = model_urls['focalnet_base_iso']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

# FocalNet large+ models 
def focalnet_large_fl3(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=192, **kwargs)
    if pretrained:
        url = model_urls['focalnet_large_fl3']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_large_fl4(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=192, **kwargs)
    if pretrained:
        url = model_urls['focalnet_large_fl4']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_xlarge_fl3(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=256, **kwargs)
    if pretrained:
        url = model_urls['focalnet_xlarge_fl3']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_xlarge_fl4(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=256, **kwargs)
    if pretrained:
        url = model_urls['focalnet_xlarge_fl4']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_huge_fl3(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=352, **kwargs)
    if pretrained:
        url = model_urls['focalnet_huge_fl3']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

def focalnet_huge_fl4(pretrained=False, **kwargs):
    model = FocalNet(depths=[2, 2, 18, 2], embed_dim=352, **kwargs)
    if pretrained:
        url = model_urls['focalnet_huge_fl4']
        checkpoint = torch.hub.load_state_dict_from_url(url=url, map_location="cpu")
        model.load_state_dict(update_weight(model.state_dict(), checkpoint["model"]))
    return model

if __name__ == '__main__':
    from copy import deepcopy
    img_size = 640
    x = torch.rand(16, 3, img_size, img_size).cuda()
    model = focalnet_tiny_srf(pretrained=True).cuda()
    # model_copy = deepcopy(model)
    for i in model(x):
        print(i.size())

    flops = model.flops()
    print(f"number of GFLOPs: {flops / 1e9}")

    n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
    print(f"number of params: {n_parameters}")
    
    print(list(model_urls.keys()))

Backbone替换

yolo.py修改

def parse_model函数

def parse_model(d, ch):  # model_dict, input_channels(3)
    # Parse a YOLOv5 model.yaml dictionary
    LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")
    anchors, nc, gd, gw, act = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'], d.get('activation')
    if act:
        Conv.default_act = eval(act)  # redefine default activation, i.e. Conv.default_act = nn.SiLU()
        LOGGER.info(f"{colorstr('activation:')} {act}")  # print
    na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchors
    no = na * (nc + 5)  # number of outputs = anchors * (classes + 5)

    is_backbone = False
    layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch out
    for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, args
        try:
            t = m
            m = eval(m) if isinstance(m, str) else m  # eval strings
        except:
            pass
        for j, a in enumerate(args):
            with contextlib.suppress(NameError):
                try:
                    args[j] = eval(a) if isinstance(a, str) else a  # eval strings
                except:
                    args[j] = a

        n = n_ = max(round(n * gd), 1) if n > 1 else n  # depth gain
        if m in {
                Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,
                BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x}:
            c1, c2 = ch[f], args[0]
            if c2 != no:  # if not output
                c2 = make_divisible(c2 * gw, 8)

            args = [c1, c2, *args[1:]]
            if m in {BottleneckCSP, C3, C3TR, C3Ghost, C3x}:
                args.insert(2, n)  # number of repeats
                n = 1
        elif m is nn.BatchNorm2d:
            args = [ch[f]]
        elif m is Concat:
            c2 = sum(ch[x] for x in f)
        # TODO: channel, gw, gd
        elif m in {Detect, Segment}:
            args.append([ch[x] for x in f])
            if isinstance(args[1], int):  # number of anchors
                args[1] = [list(range(args[1] * 2))] * len(f)
            if m is Segment:
                args[3] = make_divisible(args[3] * gw, 8)
        elif m is Contract:
            c2 = ch[f] * args[0] ** 2
        elif m is Expand:
            c2 = ch[f] // args[0] ** 2
        elif isinstance(m, str):
            t = m
            m = timm.create_model(m, pretrained=args[0], features_only=True)
            c2 = m.feature_info.channels()
        elif m in {focalnet_tiny_srf}: #可添加更多Backbone
            m = m(*args)
            c2 = m.channel
        else:
            c2 = ch[f]
        if isinstance(c2, list):
            is_backbone = True
            m_ = m
            m_.backbone = True
        else:
            m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # module
            t = str(m)[8:-2].replace('__main__.', '')  # module type
        np = sum(x.numel() for x in m_.parameters())  # number params
        m_.i, m_.f, m_.type, m_.np = i + 4 if is_backbone else i, f, t, np  # attach index, 'from' index, type, number params
        LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # print
        save.extend(x % (i + 4 if is_backbone else i) for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist
        layers.append(m_)
        if i == 0:
            ch = []
        if isinstance(c2, list):
            ch.extend(c2)
            for _ in range(5 - len(ch)):
                ch.insert(0, 0)
        else:
            ch.append(c2)
    return nn.Sequential(*layers), sorted(save)

def _forward_once函数

def _forward_once(self, x, profile=False, visualize=False):
    y, dt = [], []  # outputs
    for m in self.model:
        if m.f != -1:  # if not from previous layer
            x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
        if profile:
            self._profile_one_layer(m, x, dt)
        if hasattr(m, 'backbone'):
            x = m(x)
            for _ in range(5 - len(x)):
                x.insert(0, None)
            for i_idx, i in enumerate(x):
                if i_idx in self.save:
                    y.append(i)
                else:
                    y.append(None)
            x = x[-1]
        else:
            x = m(x)  # run
            y.append(x if m.i in self.save else None)  # save output
        if visualize:
            feature_visualization(x, m.type, m.i, save_dir=visualize)
    return x

创建新的.yaml配置文件

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license

# Parameters
nc: 80  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.25  # layer channel multiple
anchors:
  - [10,13, 16,30, 33,23]  # P3/8
  - [30,61, 62,45, 59,119]  # P4/16
  - [116,90, 156,198, 373,326]  # P5/32

# 0-P1/2
# 1-P2/4
# 2-P3/8
# 3-P4/16
# 4-P5/32

# YOLOv5 v6.0 backbone
backbone:
  # [from, number, module, args]
  [[-1, 1, focalnet_tiny_srf, [False]], # 4
   [-1, 1, SPPF, [1024, 5]],  # 5
  ]

# YOLOv5 v6.0 head
head:
  [[-1, 1, Conv, [512, 1, 1]], # 6
   [-1, 1, nn.Upsample, [None, 2, 'nearest']], # 7
   [[-1, 3], 1, Concat, [1]],  # cat backbone P4 8
   [-1, 3, C3, [512, False]],  # 9

   [-1, 1, Conv, [256, 1, 1]], # 10
   [-1, 1, nn.Upsample, [None, 2, 'nearest']], # 11
   [[-1, 2], 1, Concat, [1]],  # cat backbone P3 12
   [-1, 3, C3, [256, False]],  # 13 (P3/8-small)

   [-1, 1, Conv, [256, 3, 2]], # 14
   [[-1, 10], 1, Concat, [1]],  # cat head P4 15
   [-1, 3, C3, [512, False]],  # 16 (P4/16-medium)

   [-1, 1, Conv, [512, 3, 2]], # 17
   [[-1, 5], 1, Concat, [1]],  # cat head P5 18
   [-1, 3, C3, [1024, False]],  # 19 (P5/32-large)

   [[13, 16, 19], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
  ]