当前位置：首页 > article >正文

Windows11+PyCharm利用MMSegmentation训练自己的数据集保姆级教程

article 2025/2/12 12:12:12

系统版本：Windows 11

依赖环境：Anaconda3

运行软件：PyCharm

一.环境配置

通过Anaconda Prompt(anaconda)打开终端
创建一个虚拟环境

conda create --name mmseg python=3.9

3.激活虚拟环境

conda activate mmseg

4.安装pytorch和cuda

torch版本要求是1.12或者1.13，这里选择安装1.12，安装命令从pytorch官网找，地址

pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116

5.安装mmcv

安装命令生成地址：地址

pip install mmcv==2.0.0rc4 -f https://download.openmmlab.com/mmcv/dist/cu116/torch1.12/index.html

6.下载源码

这里可以去github上下载1.1.1版本代码，也可以下载我准备好的代码：地址

7.用pycharm打开mmsegmentation-1.1.1

选择好配置的环境之后，打开终端，运行如下命令：

pip install -v -e .

至此环境配置完成。

二.准备自己的数据集

数据集的准备请查看：数据集制作教程

上面提供的源码中包含可训练的数据集，可以直接下载!

三.开始训练

在pycharm中打开上面下载的源码。

1.在mmseg/datasets文件夹下新建mysegDataset.py

from mmseg.registry import DATASETS
from .basesegdataset import BaseSegDataset

@DATASETS.register_module()
class mysegDataset(BaseSegDataset):
    # 类别和对应的 RGB配色
    METAINFO = {
        'classes':['background', 'red', 'green', 'white', 'seed-black', 'seed-white'],
        'palette':[[127,127,127], [200,0,0], [0,200,0], [144,238,144], [30,30,30], [251,189,8]]
    }
    
    # 指定图像扩展名、标注扩展名
    def __init__(self,
                 seg_map_suffix='.png',   # 标注mask图像的格式
                 reduce_zero_label=False, # 类别ID为0的类别是否需要除去
                 **kwargs) -> None:
        super().__init__(
            seg_map_suffix=seg_map_suffix,
            reduce_zero_label=reduce_zero_label,
            **kwargs)

2.注册数据集

在`mmseg/datasets/__init__.py`中注册刚刚定义的`mysegDataset`数据集类，如下图所示，在最后添加即可

3.pipeline配置文件

在configs/_base_/datasets文件夹下新建mysegDataset_pipeline.py，并添加如下代码。



# 数据集路径
dataset_type = 'mysegDataset' # 数据集类名
data_root = 'Watermelon87_Semantic_Seg_Mask/' # 数据集路径（相对于mmsegmentation主目录）

# 输入模型的图像裁剪尺寸，一般是 128 的倍数，越小显存开销越少
crop_size = (512, 512)

# 训练预处理
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations'),
    dict(
        type='RandomResize',
        scale=(2048, 1024),
        ratio_range=(0.5, 2.0),
        keep_ratio=True),
    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
    dict(type='RandomFlip', prob=0.5),
    dict(type='PhotoMetricDistortion'),
    dict(type='PackSegInputs')
]

# 测试预处理
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='Resize', scale=(2048, 1024), keep_ratio=True),
    dict(type='LoadAnnotations'),
    dict(type='PackSegInputs')
]

# TTA后处理
img_ratios = [0.5, 0.75, 1.0, 1.25, 1.5, 1.75]
tta_pipeline = [
    dict(type='LoadImageFromFile', file_client_args=dict(backend='disk')),
    dict(
        type='TestTimeAug',
        transforms=[
            [
                dict(type='Resize', scale_factor=r, keep_ratio=True)
                for r in img_ratios
            ],
            [
                dict(type='RandomFlip', prob=0., direction='horizontal'),
                dict(type='RandomFlip', prob=1., direction='horizontal')
            ], [dict(type='LoadAnnotations')], [dict(type='PackSegInputs')]
        ])
]

# 训练 Dataloader
train_dataloader = dict(
    batch_size=2,
    num_workers=0,
    persistent_workers=True,
    sampler=dict(type='InfiniteSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(
            img_path='img_dir/train', seg_map_path='ann_dir/train'),
        pipeline=train_pipeline))

# 验证 Dataloader
val_dataloader = dict(
    batch_size=1,
    num_workers=0,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(
            img_path='img_dir/val', seg_map_path='ann_dir/val'),
        pipeline=test_pipeline))

# 测试 Dataloader
test_dataloader = val_dataloader

# 验证 Evaluator
val_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU', 'mDice', 'mFscore'])

# 测试 Evaluator
test_evaluator = val_evaluator

4.配置生成

在主目录下新建configset.py，并添加如下代码。

改代码中主要用于配置训练参数，右键运行生成配置文件。

from mmengine import Config
cfg = Config.fromfile('./configs/unet/unet-s5-d16_fcn_4xb4-160k_cityscapes-512x1024.py') ##选择训练模型
dataset_cfg = Config.fromfile('./configs/_base_/datasets/mysegDataset_pipeline.py')   ## 选择pipeline
cfg.merge_from_dict(dataset_cfg)

# 类别个数
NUM_CLASS = 6

cfg.crop_size = (256, 256)
cfg.model.data_preprocessor.size = cfg.crop_size
cfg.model.data_preprocessor.test_cfg = dict(size_divisor=128)

# 单卡训练时，需要把 SyncBN 改成 BN
cfg.norm_cfg = dict(type='BN', requires_grad=True) # 只使用GPU时，BN取代SyncBN
cfg.model.backbone.norm_cfg = cfg.norm_cfg
cfg.model.decode_head.norm_cfg = cfg.norm_cfg
cfg.model.auxiliary_head.norm_cfg = cfg.norm_cfg

# 模型 decode/auxiliary 输出头，指定为类别个数
cfg.model.decode_head.num_classes = NUM_CLASS
cfg.model.auxiliary_head.num_classes = NUM_CLASS

# 训练 Batch Size
cfg.train_dataloader.batch_size = 2

# 结果保存目录
cfg.work_dir = './work_dirs/mysegDataset-UNet'

# 模型保存与日志记录
cfg.train_cfg.max_iters = 10000 # 训练迭代次数
cfg.train_cfg.val_interval = 500 # 评估模型间隔
cfg.default_hooks.logger.interval = 100 # 日志记录间隔
cfg.default_hooks.checkpoint.interval = 2500 # 模型权重保存间隔
cfg.default_hooks.checkpoint.max_keep_ckpts = 1 # 最多保留几个模型权重
cfg.default_hooks.checkpoint.save_best = 'mIoU' # 保留指标最高的模型权重

# 随机数种子
cfg['randomness'] = dict(seed=0)

cfg.dump('myconfigs/mysegDataset_UNet.py')

5.修改num_workers=0

使用Windows系统训练，将上一步生成的配置文件中所有的num_workers修改成0。

crop_size = (
    256,
    256,
)
data_preprocessor = dict(
    bgr_to_rgb=True,
    mean=[
        123.675,
        116.28,
        103.53,
    ],
    pad_val=0,
    seg_pad_val=255,
    size=(
        512,
        1024,
    ),
    std=[
        58.395,
        57.12,
        57.375,
    ],
    type='SegDataPreProcessor')
data_root = 'Watermelon87_Semantic_Seg_Mask/'
dataset_type = 'mysegDataset'
default_hooks = dict(
    checkpoint=dict(
        by_epoch=False,
        interval=2500,
        max_keep_ckpts=1,
        save_best='mIoU',
        type='CheckpointHook'),
    logger=dict(interval=100, log_metric_by_epoch=False, type='LoggerHook'),
    param_scheduler=dict(type='ParamSchedulerHook'),
    sampler_seed=dict(type='DistSamplerSeedHook'),
    timer=dict(type='IterTimerHook'),
    visualization=dict(type='SegVisualizationHook'))
default_scope = 'mmseg'
env_cfg = dict(
    cudnn_benchmark=True,
    dist_cfg=dict(backend='nccl'),
    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
img_ratios = [
    0.5,
    0.75,
    1.0,
    1.25,
    1.5,
    1.75,
]
load_from = None
log_level = 'INFO'
log_processor = dict(by_epoch=False)
model = dict(
    auxiliary_head=dict(
        align_corners=False,
        channels=64,
        concat_input=False,
        dropout_ratio=0.1,
        in_channels=128,
        in_index=3,
        loss_decode=dict(
            loss_weight=0.4, type='CrossEntropyLoss', use_sigmoid=False),
        norm_cfg=dict(requires_grad=True, type='BN'),
        num_classes=6,
        num_convs=1,
        type='FCNHead'),
    backbone=dict(
        act_cfg=dict(type='ReLU'),
        base_channels=64,
        conv_cfg=None,
        dec_dilations=(
            1,
            1,
            1,
            1,
        ),
        dec_num_convs=(
            2,
            2,
            2,
            2,
        ),
        downsamples=(
            True,
            True,
            True,
            True,
        ),
        enc_dilations=(
            1,
            1,
            1,
            1,
            1,
        ),
        enc_num_convs=(
            2,
            2,
            2,
            2,
            2,
        ),
        in_channels=3,
        norm_cfg=dict(requires_grad=True, type='BN'),
        norm_eval=False,
        num_stages=5,
        strides=(
            1,
            1,
            1,
            1,
            1,
        ),
        type='UNet',
        upsample_cfg=dict(type='InterpConv'),
        with_cp=False),
    data_preprocessor=dict(
        bgr_to_rgb=True,
        mean=[
            123.675,
            116.28,
            103.53,
        ],
        pad_val=0,
        seg_pad_val=255,
        size=(
            256,
            256,
        ),
        std=[
            58.395,
            57.12,
            57.375,
        ],
        test_cfg=dict(size_divisor=128),
        type='SegDataPreProcessor'),
    decode_head=dict(
        align_corners=False,
        channels=64,
        concat_input=False,
        dropout_ratio=0.1,
        in_channels=64,
        in_index=4,
        loss_decode=dict(
            loss_weight=1.0, type='CrossEntropyLoss', use_sigmoid=False),
        norm_cfg=dict(requires_grad=True, type='BN'),
        num_classes=6,
        num_convs=1,
        type='FCNHead'),
    pretrained=None,
    test_cfg=dict(crop_size=256, mode='whole', stride=170),
    train_cfg=dict(),
    type='EncoderDecoder')
norm_cfg = dict(requires_grad=True, type='BN')
optim_wrapper = dict(
    clip_grad=None,
    optimizer=dict(lr=0.01, momentum=0.9, type='SGD', weight_decay=0.0005),
    type='OptimWrapper')
optimizer = dict(lr=0.01, momentum=0.9, type='SGD', weight_decay=0.0005)
param_scheduler = [
    dict(
        begin=0,
        by_epoch=False,
        end=160000,
        eta_min=0.0001,
        power=0.9,
        type='PolyLR'),
]
randomness = dict(seed=0)
resume = False
test_cfg = dict(type='TestLoop')
test_dataloader = dict(
    batch_size=1,
    dataset=dict(
        data_prefix=dict(img_path='img_dir/val', seg_map_path='ann_dir/val'),
        data_root='Watermelon87_Semantic_Seg_Mask/',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(keep_ratio=True, scale=(
                2048,
                1024,
            ), type='Resize'),
            dict(type='LoadAnnotations'),
            dict(type='PackSegInputs'),
        ],
        type='mysegDataset'),
    num_workers=0,
    persistent_workers=False,
    sampler=dict(shuffle=False, type='DefaultSampler'))
test_evaluator = dict(
    iou_metrics=[
        'mIoU',
        'mDice',
        'mFscore',
    ], type='IoUMetric')
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(keep_ratio=True, scale=(
        2048,
        1024,
    ), type='Resize'),
    dict(type='LoadAnnotations'),
    dict(type='PackSegInputs'),
]
train_cfg = dict(max_iters=10000, type='IterBasedTrainLoop', val_interval=500)
train_dataloader = dict(
    batch_size=2,
    dataset=dict(
        data_prefix=dict(
            img_path='img_dir/train', seg_map_path='ann_dir/train'),
        data_root='Watermelon87_Semantic_Seg_Mask/',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(type='LoadAnnotations'),
            dict(
                keep_ratio=True,
                ratio_range=(
                    0.5,
                    2.0,
                ),
                scale=(
                    2048,
                    1024,
                ),
                type='RandomResize'),
            dict(
                cat_max_ratio=0.75, crop_size=(
                    512,
                    512,
                ), type='RandomCrop'),
            dict(prob=0.5, type='RandomFlip'),
            dict(type='PhotoMetricDistortion'),
            dict(type='PackSegInputs'),
        ],
        type='mysegDataset'),
    num_workers=0,
    persistent_workers=False,
    sampler=dict(shuffle=True, type='InfiniteSampler'))
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations'),
    dict(
        keep_ratio=True,
        ratio_range=(
            0.5,
            2.0,
        ),
        scale=(
            2048,
            1024,
        ),
        type='RandomResize'),
    dict(cat_max_ratio=0.75, crop_size=(
        512,
        512,
    ), type='RandomCrop'),
    dict(prob=0.5, type='RandomFlip'),
    dict(type='PhotoMetricDistortion'),
    dict(type='PackSegInputs'),
]
tta_model = dict(type='SegTTAModel')
tta_pipeline = [
    dict(file_client_args=dict(backend='disk'), type='LoadImageFromFile'),
    dict(
        transforms=[
            [
                dict(keep_ratio=True, scale_factor=0.5, type='Resize'),
                dict(keep_ratio=True, scale_factor=0.75, type='Resize'),
                dict(keep_ratio=True, scale_factor=1.0, type='Resize'),
                dict(keep_ratio=True, scale_factor=1.25, type='Resize'),
                dict(keep_ratio=True, scale_factor=1.5, type='Resize'),
                dict(keep_ratio=True, scale_factor=1.75, type='Resize'),
            ],
            [
                dict(direction='horizontal', prob=0.0, type='RandomFlip'),
                dict(direction='horizontal', prob=1.0, type='RandomFlip'),
            ],
            [
                dict(type='LoadAnnotations'),
            ],
            [
                dict(type='PackSegInputs'),
            ],
        ],
        type='TestTimeAug'),
]
val_cfg = dict(type='ValLoop')
val_dataloader = dict(
    batch_size=1,
    dataset=dict(
        data_prefix=dict(img_path='img_dir/val', seg_map_path='ann_dir/val'),
        data_root='Watermelon87_Semantic_Seg_Mask/',
        pipeline=[
            dict(type='LoadImageFromFile'),
            dict(keep_ratio=True, scale=(
                2048,
                1024,
            ), type='Resize'),
            dict(type='LoadAnnotations'),
            dict(type='PackSegInputs'),
        ],
        type='mysegDataset'),
    num_workers=0,
    persistent_workers=False,
    sampler=dict(shuffle=False, type='DefaultSampler'))
val_evaluator = dict(
    iou_metrics=[
        'mIoU',
        'mDice',
        'mFscore',
    ], type='IoUMetric')
vis_backends = [
    dict(type='LocalVisBackend'),
]
visualizer = dict(
    name='visualizer',
    type='SegLocalVisualizer',
    vis_backends=[
        dict(type='LocalVisBackend'),
    ])
work_dir = './work_dirs/mysegDataset-UNet'

6.训练模型

在终端中运行以下命令：

 python tools/train.py myconfigs/mysegDataset_UNet.py

查看全文

http://www.kler.cn/a/542041.html

【R语言】相关系数

MYSQL利用PXC实现高可用

在CT107D单片机综合训练平台上实现外部中断控制LED闪烁

强化学习之 PPO 算法：原理、实现与案例深度剖析

MongoDB 有哪些特性

[RabbitMQ] RabbitMQ常见面试题

使用 Visual Studio Code (VS Code) 开发 Python 图形界面程序

Day59_20250207_图论part4_110.字符串接龙|105.有向图的完全可达性|106.岛屿的周长

Spring Boot整合DeepSeek实现AI对话（API调用和本地部署）

淘宝App交易链路终端混合场景体验探索

教育局网络设备运维和资产管理方案

SpringBoot中能被外部注入以来的注解

网站快速收录攻略：提升页面加载速度

django中间件，中间件给下面传值

05：定时器生成频率不同的波形

Rocketmq 和 Rabbitmq ，在多消费者的情况下，可以实现顺序消费吗

使用腾讯云大模型知识引擎搭建满血deepseek

arcgis for js实现层叠立体效果

C++ 用Eigen的非线性求解LevenbergMarquardt，亲测ok

Python 透明数字时钟

独立站赋能反向海淘：跨境代购系统的用户体验与支付解决方案

2025.2.10 每日学习记录3：技术报告只差相关工作+补实验

在npm上传属于自己的包

【RabbitMQ的x-death头】消息死亡记录头流转示例

攻防世界32 very_easy_sql【SSRF/SQL时间盲注】

android的第一个app项目（java版）

一.环境配置

二.准备自己的数据集

三.开始训练

相关文章：