VMambe应用目标检测backbone之VSSM结构打印

已于 2024-07-25 09:21:04 修改
阅读量298 收藏 1
点赞数 8
文章标签： 目标检测 人工智能 计算机视觉
于 2024-07-25 09:19:52 首次发布
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/qq_47659236/article/details/140679968
版权

[Releases · MzeroMiko/VMamba · GitHub](https://github.com/MzeroMiko/VMamba/releases)

安装torchvision等：

尝试版本1：wins上python=3.10, torch=1.13.1  成功

    pip install torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117

尝试版本2 ：wins上python=3.10, torch=2.1.0

  pip install torchvision==0.16.0+cu118 torchaudio==2.1.0 --extra-index-url https://download.pytorch.org/whl/cu118

模型pipline:

熟悉模型结构：

import os
from functools import partial
from typing import Callable

import torch
from torch import nn
from torch.utils import checkpoint

from mmengine.model import BaseModule
from mmdet.registry import MODELS as MODELS_MMDET
from mmseg.registry import MODELS as MODELS_MMSEG

def import_abspy(name="models", path="classification/"):
    import sys
    import importlib
    path = os.path.abspath(path)
    assert os.path.isdir(path)
    sys.path.insert(0, path)
    module = importlib.import_module(name)
    sys.path.pop(0)
    #打印
    print("detection imported module: {}".format(module.__name__))
    return module

def main():
    build = import_abspy(
        "models",
        os.path.join(os.path.dirname(os.path.abspath(__file__)), "../classification/"),
    )

    Backbone_VSSM = build.vmamba.Backbone_VSSM

    class MM_VSSM(BaseModule, Backbone_VSSM):
        def __init__(self, *args, **kwargs):
            BaseModule.__init__(self)
            Backbone_VSSM.__init__(self, *args, **kwargs)

    # 测试初始化 MM_VSSM 类
    model = MM_VSSM()

    # 打印模型的类名，确保导入和初始化正确
    print("Model class:", model.__class__.__name__)
    print("Model:", model)
    print(model.__dict__)


if __name__ == "__main__":
    main()

print("Model:", model)

Model: MM_VSSM( 
  (patch_embed): Sequential(                                                                                          VMamba1427
    (0): Conv2d(3, 96, kernel_size=(4, 4), stride=(4, 4))
    (1): Permute()                   
    (2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
  )
  (layers): ModuleList(
    (0): Sequential(
      (blocks): Sequential(
        (0): VSSBlock(
          (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=96, out_features=384, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
            (out_act): Identity()
            (out_proj): Linear(in_features=192, out_features=96, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.0)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (1): VSSBlock(
          (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=96, out_features=384, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
            (out_act): Identity()
            (out_proj): Linear(in_features=192, out_features=96, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.0071428571827709675)
          (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=96, out_features=384, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=384, out_features=96, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
      )
      (downsample): Sequential(
        (0): Permute()
        (1): Conv2d(96, 192, kernel_size=(2, 2), stride=(2, 2))
        (2): Permute()
        (3): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
      )
    )
    (1): Sequential(
      (blocks): Sequential(
        (0): VSSBlock(
          (norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=192, out_features=768, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
            (out_act): Identity()
            (out_proj): Linear(in_features=384, out_features=192, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.014285714365541935)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (1): VSSBlock(
          (norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=192, out_features=768, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
            (out_act): Identity()
            (out_proj): Linear(in_features=384, out_features=192, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.02142857201397419)
          (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=192, out_features=768, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=768, out_features=192, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
      )
      (downsample): Sequential(
        (0): Permute()
        (1): Conv2d(192, 384, kernel_size=(2, 2), stride=(2, 2))
        (2): Permute()
        (3): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
      )
    )
    (2): Sequential(
      (blocks): Sequential(
        (0): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.02857142873108387)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (1): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.0357142873108387)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (2): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.04285714402794838)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (3): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.05000000074505806)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (4): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.05714285746216774)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (5): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.06428571045398712)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (6): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.0714285746216774)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (7): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.07857143133878708)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (8): VSSBlock(
          (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=384, out_features=1536, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
            (out_act): Identity()
            (out_proj): Linear(in_features=768, out_features=384, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.08571428805589676)
          (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=1536, out_features=384, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
      )
      (downsample): Sequential(
        (0): Permute()
        (1): Conv2d(384, 768, kernel_size=(2, 2), stride=(2, 2))
        (2): Permute()
        (3): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
      )
    )
    (3): Sequential(
      (blocks): Sequential(
        (0): VSSBlock(
          (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=768, out_features=3072, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1536)
            (out_act): Identity()
            (out_proj): Linear(in_features=1536, out_features=768, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.09285714477300644)
          (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=768, out_features=3072, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=3072, out_features=768, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
        (1): VSSBlock(
          (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (op): SS2D(
            (out_norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
            (in_proj): Linear(in_features=768, out_features=3072, bias=False)
            (act): SiLU()
            (conv2d): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1536)
            (out_act): Identity()
            (out_proj): Linear(in_features=1536, out_features=768, bias=False)
            (dropout): Identity()
          )
          (drop_path): timm.DropPath(0.10000000149011612)
          (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (mlp): Mlp(
            (fc1): Linear(in_features=768, out_features=3072, bias=True)
            (act): GELU(approximate='none')
            (fc2): Linear(in_features=3072, out_features=768, bias=True)
            (drop): Dropout(p=0.0, inplace=False)
          )
        )
      )
      (downsample): Identity()
    )
  )
  (outnorm0): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
  (outnorm1): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
  (outnorm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
  (outnorm3): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
)

模型参数

print(model.__dict__)

`__dict__` 是一个字典（`dict`），它包含了对象的所有属性名（作为键）和对应的值。

对于你的 `MM_VSSM` 类的实例，如果这个类没有定义任何特殊的属性，并且没有在实例化后添加任何属性，那么 `model.__dict__` 可能不会显示任何内容，因为它可能只包含一些由 Python 在创建实例时自动添加的属性，比如 `__module__`、`__doc__` 等。

{'training': True, '_parameters': OrderedDict(), '_buffers': OrderedDict(), '_non_persistent_buffers_set': set(), '_backward_pre_hooks': OrderedDict(), '_backward_hooks': OrderedDict(), '_is_full_backward_hook': None, '_forward_hooks': OrderedDict(), '_forward_hooks_with_kwargs': OrderedDict(), '_forward_hooks_always_called': OrderedDict(), '_forward_pre_hooks': OrderedDict(), '_forward_pre_hooks_with_kwargs': OrderedDict(), '_state_dict_hooks': OrderedDict(), '_state_dict_pre_hooks': OrderedDict(), '_load_state_dict_pre_hooks': OrderedDict(), '_load_state_dict_post_hooks': OrderedDict(), '_modules': OrderedDict([('patch_embed', Sequential(
  (0): Conv2d(3, 96, kernel_size=(4, 4), stride=(4, 4))
  (1): Permute()
  (2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
)), ('layers', ModuleList(
  (0): Sequential(
    (blocks): Sequential(
      (0): VSSBlock(
        (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=96, out_features=384, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (out_act): Identity()
          (out_proj): Linear(in_features=192, out_features=96, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.0)
        (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=96, out_features=384, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=384, out_features=96, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (1): VSSBlock(
        (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=96, out_features=384, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(192, 192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=192)
          (out_act): Identity()
          (out_proj): Linear(in_features=192, out_features=96, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.0071428571827709675)
        (norm2): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=96, out_features=384, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=384, out_features=96, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
    )
    (downsample): Sequential(
      (0): Permute()
      (1): Conv2d(96, 192, kernel_size=(2, 2), stride=(2, 2))
      (2): Permute()
      (3): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
    )
  )
  (1): Sequential(
    (blocks): Sequential(
      (0): VSSBlock(
        (norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=192, out_features=768, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (out_act): Identity()
          (out_proj): Linear(in_features=384, out_features=192, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.014285714365541935)
        (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=192, out_features=768, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=768, out_features=192, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (1): VSSBlock(
        (norm): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=192, out_features=768, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=384)
          (out_act): Identity()
          (out_proj): Linear(in_features=384, out_features=192, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.02142857201397419)
        (norm2): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=192, out_features=768, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=768, out_features=192, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
    )
    (downsample): Sequential(
      (0): Permute()
      (1): Conv2d(192, 384, kernel_size=(2, 2), stride=(2, 2))
      (2): Permute()
      (3): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
    )
  )
  (2): Sequential(
    (blocks): Sequential(
      (0): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.02857142873108387)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (1): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.0357142873108387)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (2): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.04285714402794838)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (3): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.05000000074505806)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (4): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.05714285746216774)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (5): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.06428571045398712)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (6): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.0714285746216774)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (7): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.07857143133878708)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (8): VSSBlock(
        (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=384, out_features=1536, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(768, 768, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=768)
          (out_act): Identity()
          (out_proj): Linear(in_features=768, out_features=384, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.08571428805589676)
        (norm2): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=384, out_features=1536, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=1536, out_features=384, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
    )
    (downsample): Sequential(
      (0): Permute()
      (1): Conv2d(384, 768, kernel_size=(2, 2), stride=(2, 2))
      (2): Permute()
      (3): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
    )
  )
  (3): Sequential(
    (blocks): Sequential(
      (0): VSSBlock(
        (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=768, out_features=3072, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1536)
          (out_act): Identity()
          (out_proj): Linear(in_features=1536, out_features=768, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.09285714477300644)
        (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=768, out_features=3072, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=3072, out_features=768, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
      (1): VSSBlock(
        (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
        (op): SS2D(
          (out_norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
          (in_proj): Linear(in_features=768, out_features=3072, bias=False)
          (act): SiLU()
          (conv2d): Conv2d(1536, 1536, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=1536)
          (out_act): Identity()
          (out_proj): Linear(in_features=1536, out_features=768, bias=False)
          (dropout): Identity()
        )
        (drop_path): timm.DropPath(0.10000000149011612)
        (norm2): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
        (mlp): Mlp(
          (fc1): Linear(in_features=768, out_features=3072, bias=True)
          (act): GELU(approximate='none')
          (fc2): Linear(in_features=3072, out_features=768, bias=True)
          (drop): Dropout(p=0.0, inplace=False)
        )
      )
    )
    (downsample): Identity()
  )
)), ('outnorm0', LayerNorm((96,), eps=1e-05, elementwise_affine=True)), ('outnorm1', LayerNorm((192,), eps=1e-05, elementwise_affine=True)), ('outnorm2', LayerNorm((384,), eps=1e-05, elementwise_affine=True)), ('outnorm3', LayerNorm((768,), eps=1e-05, elementwise_affine=True))]), 'channel_first': False, 'num_classes': 1000, 'num_layers': 4, 'num_features': 768, 'dims': [96, 192, 384, 768], 'pos_embed': None, 'out_indices': (0, 1, 2, 3), '_is_init': False, 'init_cfg': None}

放在心尖L
关注
  • 8
    点赞
  • 踩
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Swin-UMamba—基于 Mamba 的 UNet 和基于 ImageNet 的预训练—论文精读和代码实践
Philo的博客
03-06 3061
准确的医学图像分割需要集成多尺度信息从局部特征到全局依赖性。然而，现有方法对远程全局信息进行建模具有挑战性，其中卷积神经网络（CNN）受到其局部感受野的限制，而视觉变换器（ViT）则受到其注意力机制的高二次复杂度的影响。最近，基于 Mamba 的模型因其在长序列建模方面令人印象深刻的能力而受到极大关注。多项研究表明，这些模型在各种任务中都可以优于流行的视觉模型，提供更高的准确性、更低的内存消耗和更少的计算负担。
深度解析：YOLOv5中的CSP结构及其在Backbone中的应用
06-29
2. **网络结构**：YOLOv5的网络结构包括输入端、Backbone骨干网络、Neck颈部结构和Head头部结构（Prediction），分别负责输入图片预处理、特征提取、特征融合、输出检测信息。 3. **数据增强**：YOLOv5采用了Mosaic...
VSSM VMamba实现
weixin_45668967的博客
03-14 3551
图中的SS2D和SS2D类的定义有偏差，简单来说是是包含SS2D块加一个残差连接，图中所示SS2D应表示状态空间模型SSM部分，即VSS块相比SS2D块只增加了残差连接和入口的归一化。对于ssm分支来说，其输入输出维度不变为(B, H, W, d_model) ，对于mlp分支来说中间的隐藏维度根据mlp_ratio参数定义会有所增加，但是最后又会映射为原来的维度，因此整体上并不改变输入的维度。下采样模块，通过2D卷积之后，长宽变为原来的一半，通道数不变。，嵌入后通道维变为96，
Mamba又杀入ECCV 2024！清华提出MambaIR：图像复原超强SOTA模型
阿木寺的博客
07-03 404
点击下方卡片，关注“CVer”公众号AI/CV重磅干货，第一时间送达点击进入—>【Mamba/多模态/扩散】交流群添加微信：CVer5555，小助手会拉你进群！扫描下方二维码，加入CVer学术星球！可以获得最新顶会/顶刊上的论文idea和CV从入门到精通资料，及最前沿应用！发论文/搞科研/涨薪，强烈推荐！转载自：极市平台 作者：一路向北来源丨https://zhuanlan.zhihu....
目标检测中的BackBone,Neck,Head
lzz555517的博客
01-10 3745
目标检测中的BackBone,Neck,Head 本博客只记录本人学习对目标检测的理解，如有不正确，烦请各位大佬指正。 基于深度学习的目标检测算法有三个组件：Backbone,Neck及Head。 首先，对于backbone，看文献或博客经常会有backbone，看翻译会译作主干，其实就是主干网络，主要用于特征提取，并且基本在大型数据集(ImageNet,COCO,VOC等)上完成训练，拥有预训练参数的卷积神经网络，例如：ResNet50，Darknet53等等。 然后Head,翻译为头，检测头，主要用于预
YOLOv4目标检测-Backbone
一个努力写代码的小菜鸟，目前主要从事图像算法，目标检测/分割，模型量化与部署的工作
02-12 3359
YOLOv4源码详解笔记---Backbone---CSPDarknet53
Swin Transformer做backbone的YoloX目标检测
shayinzzh的博客
05-30 1万+
使用Swin Transformer做骨干网络进行YoloX目标检测
目标检测（Object Detection）
热门推荐
yegeli的博客
11-20 21万+
文章目录目标检测（Object Detection）一、基本概念1. 什么是目标检测2. 目标检测的核心问题3. 目标检测算法分类1）Tow Stage2）One Stage4. 目标检测应用1）人脸检测2）行人检测3）车辆检测4）遥感检测二、目标检测原理1. 候选区域产生1）滑动窗口2）选择性搜索① 什么是选择性搜索② 选择搜索流程③ 选择搜索优点2. 数据表示3. 效果评估4. 非极大值抑制三、目标检测模型1. R-CNN系列1）R-CNN① 定义② 流程③ 效果④ 缺点2）Fast R-CNN① 定义
目标检测算法中的Backbone、Neck和Head
RayChiu757374816的博客
02-24 4363
yolov4论文中对目标检测中Backbone、Neck和Head三个术语的解释
睿智的目标检测66——Pytorch搭建YoloV8目标检测平台
Bubbliiiing的学习小课堂
05-15 4万+
又搞了个YoloV8，看起来似乎在抢这个名字。YoloV8所使用的主干特征提取网络主要为速度快做了一些优化：1、颈部结构使用普通的步长为2的3x3卷积。YoloV5最初使用了Focus结构来初步提取特征，在改进后使用了大卷积核的卷积来初步提取特征，速度都不快。YoloV7则使用了三次卷积来初步提取特征，速度也不快。YoloV8则使用普通的步长为2的3x3卷积核来初步提取特征（估计是感受野够了）。这样做会损失一些感受野，但是可以提高模型的速度。
目标检测算法——YOLOV7——详解
TigerZ*的博客
08-02 17万+
本文详细解读了YOLOV7 0.1版本代码的网络结构、损失函数、正负样本匹配规则等。可以高效的快速掌握YOLOV7的宏观结构和真正实现时的核心点。
YOLOX-结合YOLOX+SwinTransformer作为Backbone实现的目标检测算法-附预训练权重下载+项目源码
05-09
在实际应用中，这样的目标检测系统可以广泛应用于视频监控、自动驾驶、无人机侦查、工业质检等领域。通过预训练权重，开发者可以节省大量的训练时间和计算资源，直接在自己的应用中部署高精度的检测模型。 总结来说...
YOLOV5 改进实战项目【更换backbone为MobileNet】：大型海上图像目标检测数据集
03-19
基于YOLOV5【更换backbone为MobileNet】对大型海上目标检测数据集的目标检测实战项目，包含代码、数据集、训练好的权重参数，经测试，代码可以直接使用 【数据集介绍】海上目标图像数据，类别：邮轮、小鸟、鱼、...
目标检测Faster-RCNN源码
05-31
介绍利用Faster-RCNN去训练Pascol VOC数据集或者训练自定义数据集。 参考博客：https://blog.csdn.net/weixin_38346042/article/details/125044860?spm=1001.2014.3001.5502
##YOLOv10目标检测算法的详细概述
05-25
1. **高速度与高效率**：YOLOv10在保持较高检测速度的同时，还实现了更好的检测精度，特别适合于实时目标检测的应用场景。 2. **精确度提升**：相较于YOLOv4，YOLOv10在保持高速度的同时进一步提高了检测的准确性，...
大模型微调工具-torchtune
weixin_40777649的博客
08-08 619
1.定义2.安装3. 案例。
【OpenCV C++20 学习笔记】给图片加边框-copyMakeBorder
TeamLee的博客
08-06 529
简单介绍在OpenCV中如何用copyMakeBorder的方法给图片添加边框
根据哈希值或者特征值判断图片是否相同
最新发布
Y_warmth
08-09 394
detectAndCompute 方法结合了特征检测和描述符计算两个步骤，可以在图像中检测出特征点并计算这些特征点的描述符。ORB（Oriented FAST and Rotated BRIEF）是一种结合了 FAST 特征检测器和 BRIEF 描述符的特征提取算法，具有旋转不变性和尺度不变性，适用于图像特征匹配、目标检测等任务。接着使用 ORB 对象的 detectAndCompute() 方法检测图像中的特征点并计算描述符，最后将特征点绘制在图像上并显示出来。
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值