系统版本：Windows 10 企业版
依赖环境：Anaconda3
运行软件：PyCharm
MMSegmentation版本：V1.1.1
前提：运行环境已经配置好，环境的配置可以参考：Win10系统下MMSegmentation的环境配置-CSDN博客
目录
1. 从官网下载相应的MMSegmentation
从官网下载对应的版本：
这里可以看到不同的版本。本教程使用的是最新的V1.1.1版本。
把下载好的版本当成一个工程项目，直接在Pycharm中打开。
2. 定义数据集类
在mmseg/datasets中新建一个以自己的数据集命名的py文件，例如，我新建了一个名为cag.py的文件。
在cag.py中为自己的数据集创建一个新的类：
# Copyright (c) Lisa. All rights reserved.
"""
@Title: 创建一个自己的数据集
@Author: Lisa
@Date: 2023/09/13
"""
import mmengine.fileio as fileio
from mmseg.registry import DATASETS
from .basesegdataset import BaseSegDataset
@DATASETS.register_module()
class CAGDataset(BaseSegDataset):
"""CAG dataset.
In segmentation map annotation for CAG, 0 stands for background, which is
included in 2 categories. ``reduce_zero_label`` is fixed to False. The
``img_suffix`` is fixed to '.png' and ``seg_map_suffix`` is fixed to
'.png'.
"""
# 类别和对应的RGB配色
METAINFO = dict(
classes=('background', 'vessel'), # 类别标签名称
palette=[[0, 0, 0], [255, 255, 255]]) # 类别标签上色的RGB颜色
# 指定图像扩展名，标注扩展名
def __init__(self,
img_suffix='.png', # 输入image的后缀名为’.png‘
seg_map_suffix='.png', # 输入mask/label的后缀名为’.png‘
reduce_zero_label=False,
**kwargs) -> None:
super().__init__(
img_suffix=img_suffix,
seg_map_suffix=seg_map_suffix,
reduce_zero_label=reduce_zero_label,
**kwargs)
3. 注册数据集类
打开mmseg/datasets/__init__.py文件
导入定义好的CAGDataset类，然后添加到__all__中。
4. 配置数据处理pipeline文件
在configs/_base_/datasets中创建一个数据处理pipeline的py文件。
cag_pipeline.py中数据处理pipeline如下：
# 数据处理pipeline
# 参照同济张子豪
# dataset settings 设置数据集路径
dataset_type = 'CAGDataset' # must be the same name of custom dataset. 必须和自定义数据集名称完全一致。
data_root = '../data/CAG' # 数据集根目录， 后续所有的pipeline使用的目录都会在此目录下的子目录读取
# img_scale = (2336, 3504)
# img_scale = (512, 512)
# crop_size = (256, 256)
# 输入模型的图像裁剪尺寸，一般是128的倍数。
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='Resize', img_scale=img_scale),
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),
# add loading annotation after ``Resize`` because ground truth
# does not need to do resize data transform
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', backend_args=None),
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=4, # 4
num_workers=2, # 4 dataloader的线程数目，一般设为2， 4， 8，根据CPU核数确定，或使用os.cpu_count()函数代替，一般num_workers>=4速度提升就不再明显。
persistent_workers=True, # 一种加速图片加载的操作
sampler=dict(type='InfiniteSampler', shuffle=True), # shuffle=True是打乱图片
dataset=dict(
type=dataset_type,
data_root=data_root,
data_prefix=dict(
img_path='images/training',
seg_map_path='annotations/training'),
# ann_file='splits/train.txt',
pipeline=train_pipeline
)
)
# 测试DataLoader
val_dataloader = dict(
batch_size=1,
num_workers=4, # 4
persistent_workers=True,
sampler=dict(type='DefaultSampler', shuffle=False),
dataset=dict(
type=dataset_type,
data_root=data_root,
data_prefix=dict(
img_path='images/validation',
seg_map_path='annotations/validation'),
# ann_file='splits/val.txt',
pipeline=test_pipeline
))
# 验证DataLoader
test_dataloader = val_dataloader
# 验证 Evaluator
val_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU', 'mDice', 'mFscore']) # 分割指标评估器
test_evaluator = val_evaluator
5. 配置Config文件
以UNet为例，配置Config文件
"""
create_Unet_config.py
配置Unet的Config文件
"""
from mmengine import Config
cfg = Config.fromfile('../configs/unet/unet-s5-d16_fcn_4xb4-160k_cityscapes-512x1024.py ')
dataset_cfg = Config.fromfile('../configs/_base_/datasets/cag_pipeline.py')
cfg.merge_from_dict(dataset_cfg)
# 修改Config配置文件
# 类别个数
NUM_CLASS = 2
cfg.crop_size = (512, 512)
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改为BN
cfg.norm_cfg = dict(type='BN', requires_grad=True)
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 = 4
# 结果保存目录
cfg.work_dir = '../work_dirs/my_Unet'
# 模型保存与日志记录
cfg.train_cfg.max_iters = 60000 # 训练迭代次数
cfg.train_cfg.val_interval = 500 # 评估模型间隔 500
cfg.default_hooks.logger.interval = 100 # 日志记录间隔
cfg.default_hooks.checkpoint.interval = 500 # 模型权重保存间隔 2500
cfg.default_hooks.checkpoint.max_keep_ckpts = 2 # 最多保留几个模型权重 1， 2
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)
# 查看完整的Config配置文件
print(cfg.pretty_text)
# 保存最终的config配置文件
cfg.dump('../my_Configs/my_Unet_20230913.py')
6. 训练模型
在PyCharm中运行my_train.py主要有两种方式，
方式一：在右上方的my_train处单击“Edit Configurations...”
在Parameters:一栏中输入命令“ ../my_Configs/my_PSPNet_20230906.py ../work_dirs/my_PSPNet”，然后点击“运行”。
方式二： 改写超参数
# Copyright (c) OpenMMLab. All rights reserved.
"""
参考train.py改写
"""
import argparse
import logging
import os
import os.path as osp
import torch.backends.cudnn
from mmengine.config import Config, DictAction
from mmengine.logging import print_log
from mmengine.runner import Runner
from mmseg.registry import RUNNERS
os.environ['CUDA_LAUNCH_BLOCKING'] = '1' #（上面报错的最后一行的提示信息）
# torch.backends.cudnn.benchmark = True
# 用于解决报错：RuntimeError: CUDA error: an illegal memory access was encountered
# CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
# For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
def parse_args():
parser = argparse.ArgumentParser(description='Train a segmentor')
parser.add_argument('--config', default='../my_Configs/my_PSPNet_20230906.py', help='train config file path') # 不加--号是表示该参数是必须的
# parser.add_argument('--config', default='../configs/deeplabv3plus/deeplabv3plus_r50-d8_4xb4-80k_cag-512x512_my.py',
# help='train config file path') # 不加--号是表示该参数是必须的
# parser.add_argument('--config', default='../my_Configs/my_KNet_20230830.py',
# help='train config file path') # 不加--号是表示该参数是必须的
# parser.add_argument('--work-dir', default='work_dir', help='the dir to save logs and models')
parser.add_argument('--work-dir', default='../work_dirs/my_PSPNet', help='the dir to save logs and models')
parser.add_argument(
'--resume',
action='store_true',
default=True,
help='resume from the latest checkpoint in the work_dir automatically')
parser.add_argument(
'--amp',
action='store_true',
default=False,
help='enable automatic-mixed-precision training')
parser.add_argument(
'--cfg-options',
nargs='+',
action=DictAction,
help='override some settings in the used config, the key-value pair '
'in xxx=yyy format will be merged into config file. If the value to '
'be overwritten is a list, it should be like key="[a,b]" or key=a,b '
'It also allows nested list/tuple values, e.g. key="[(a,b),(c,d)]" '
'Note that the quotation marks are necessary and that no white space '
'is allowed.')
parser.add_argument(
'--launcher',
choices=['none', 'pytorch', 'slurm', 'mpi'],
default='none',
help='job launcher')
# When using PyTorch version >= 2.0.0, the `torch.distributed.launch`
# will pass the `--local-rank` parameter to `tools/train.py` instead
# of `--local_rank`.
parser.add_argument('--local_rank', '--local-rank', type=int, default=0)
args = parser.parse_args()
if 'LOCAL_RANK' not in os.environ:
os.environ['LOCAL_RANK'] = str(args.local_rank)
return args
def main():
args = parse_args()
# load config
cfg = Config.fromfile(args.config)
cfg.launcher = args.launcher
if args.cfg_options is not None:
cfg.merge_from_dict(args.cfg_options)
# work_dir is determined in this priority: CLI > segment in file > filename
if args.work_dir is not None:
# update configs according to CLI args if args.work_dir is not None
cfg.work_dir = args.work_dir
elif cfg.get('work_dir', None) is None:
# use config filename as default work_dir if cfg.work_dir is None
cfg.work_dir = osp.join('./work_dirs',
osp.splitext(osp.basename(args.config))[0])
# enable automatic-mixed-precision training
if args.amp is True:
optim_wrapper = cfg.optim_wrapper.type
if optim_wrapper == 'AmpOptimWrapper':
print_log(
'AMP training is already enabled in your config.',
logger='current',
level=logging.WARNING)
else:
assert optim_wrapper == 'OptimWrapper', (
'`--amp` is only supported when the optimizer wrapper type is '
f'`OptimWrapper` but got {optim_wrapper}.')
cfg.optim_wrapper.type = 'AmpOptimWrapper'
cfg.optim_wrapper.loss_scale = 'dynamic'
# resume training
cfg.resume = args.resume
# build the runner from config
if 'runner_type' not in cfg:
# build the default runner
runner = Runner.from_cfg(cfg)
else:
# build customized runner from the registry
# if 'runner_type' is set in the cfg
runner = RUNNERS.build(cfg)
# start training
runner.train()
if __name__ == '__main__':
main()
7. 测试模型
单张测试的话，可以参考以下代码：
"""
single_image_predict.py
用训练得到的模型进行预测-单张图像
"""
import numpy as np
import matplotlib.pyplot as plt
from mmseg.apis import init_model, inference_model, show_result_pyplot
import cv2
# 载入模型 KNet
# 模型 config 配置文件
config_file = '../work_dirs/my_FastSCNN/my_FastSCNN_20230911.py'
# 模型checkpoint权重文件
checkpoint_file = '../work_dirs/my_FastSCNN/best_mIoU_iter_50500.pth'
device = 'cuda:0'
model = init_model(config_file, checkpoint_file, device=device)
# 载入测试集图像或新图像
img_path = '../data/.../xxx.png'
img_bgr = cv2.imread(img_path)
# 显示原图
plt.figure(figsize=(8, 8))
plt.imshow(img_bgr[:, :, ::-1])
plt.show()
# 语义分割预测
result = inference_model(model, img_bgr)
print(result.keys())
pred_mask = result.pred_sem_seg.data[0].cpu().numpy()
print(pred_mask.shape)
print(np.unique(pred_mask))
# #****** 可视化语义分割预测结果——方法一（直接显示分割结果）******#
# 定性
plt.figure(figsize=(8, 8))
plt.imshow(pred_mask)
plt.savefig('test_result/k1-0.jpg')
plt.show()
# 定量
print(result.seg_logits.data.shape)
# #****** 可视化语义分割预测结果--方法二（叠加在原因上进行显示）******#
# 显示语义分割结果
plt.figure(figsize=(10, 8))
plt.imshow(img_bgr[:, :, ::-1])
plt.imshow(pred_mask, alpha=0.55)
plt.axis('off')
plt.savefig('test_result/k1-1.jpg')
plt.show()
# #****** 可视化语义分割预测结果--方法三(和原图并排显示) ******#
plt.figure(figsize=(14, 8))
plt.subplot(1, 2, 1)
plt.imshow(img_bgr[:, :, ::-1])
plt.axis('off')
plt.subplot(1, 2, 2)
plt.imshow(img_bgr[:, :, ::-1])
plt.imshow(pred_mask, alpha=0.6)
plt.axis('off')
plt.savefig('test_result/k1-2.jpg')
plt.show()
# #****** 可视化语义分割预测结果-方法四(按配色方案叠加在原图上显示) ******#
# 各类别的配色方案(BGR)
palette = [
['background', [127, 127, 127]],
['vessel', [0, 0, 200]]
]
palette_dict = {}
for idx, each in enumerate(palette):
palette_dict[idx] = each[1]
print('palette_dict:', palette_dict)
opacity = 0.3 # 透明度,越大越接近原图
# 将预测的整数ID,映射为对应类别的颜色
pred_mask_bgr = np.zeros((pred_mask.shape[0], pred_mask.shape[1], 3))
for idx in palette_dict.keys():
pred_mask_bgr[np.where(pred_mask == idx)] = palette_dict[idx]
pred_mask_bgr = pred_mask_bgr.astype('uint8')
# 将语义分割预测图和原图叠加显示
pred_viz = cv2.addWeighted(img_bgr, opacity, pred_mask_bgr, 1-opacity, 0)
cv2.imwrite('test_result/k1-3.jpg', pred_viz)
plt.figure(figsize=(8, 8))
plt.imshow(pred_viz[:, :, ::-1])
plt.show()
# #***** 可视化语义分割预测结果-方法五(按mmseg/datasets/cag.py里定义的类别颜色可视化) ***** #
img_viz = show_result_pyplot(model, img_path, result, opacity=0.8, title='MMSeg', out_file='test_result/k1-4.jpg')
print('the shape of img_viz:', img_viz.shape)
plt.figure(figsize=(14, 8))
plt.imshow(img_viz)
plt.show()
# #***** 可视化语义分割预测结果--方法六(加图例) ***** #
from mmseg.datasets import CAGDataset
import numpy as np
import mmcv
from PIL import Image
# 获取类别名和调色板
classes = CAGDataset.METAINFO['classes']
palette = CAGDataset.METAINFO['palette']
opacity = 0.15
# 将分割图按调色板染色
seg_map = pred_mask.astype('uint8')
seg_img = Image.fromarray(seg_map).convert('P')
seg_img.putpalette(np.array(palette, dtype=np.uint8))
from matplotlib import pyplot as plt
import matplotlib.patches as mpatches
plt.figure(figsize=(14, 8))
img_plot = ((np.array(seg_img.convert('RGB')))*(1-opacity) + mmcv.imread(img_path)*opacity) / 255
im = plt.imshow(img_plot)
# 为每一种颜色创建一个图例
patches = [mpatches.Patch(color=np.array(palette[i])/255, label=classes[i]) for i in range(len(classes))]
plt.legend(handles=patches, bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0., fontsize='large')
plt.savefig('test_result/k1-6.jpg')
plt.show()
单张测试或批量测试，也可以参考以下代码：
"""
@title: 根据配置(Config)文件和已训练好的参数(pth)进行推理
@Date: 2023/09/14
"""
# 导入必要的库
import os
import numpy as np
import cv2
from tqdm import tqdm
from mmseg.apis import init_model, inference_model, show_result_pyplot
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
from mmseg.datasets import CAGDataset
import numpy as np
import mmcv
from PIL import Image
def predict_single_img(img_path, model, save=False, save_path=None, save_model=1, show=False, show_model=1):
# 读取图像
img_bgr = cv2.imread(img_path)
# 语义分割预测
result = inference_model(model, img_bgr)
pred_mask = result.pred_sem_seg.data[0].cpu().numpy()
# print(pred_mask.shape)
# print(np.unique(pred_mask))
# 将预测的整数ID, 映射为对应类别的颜色
pred_mask_bgr = np.zeros((pred_mask.shape[0], pred_mask.shape[1], 3))
for idx in palette_dict.keys():
pred_mask_bgr[np.where(pred_mask == idx)] = palette_dict[idx]
pred_mask_bgr = pred_mask_bgr.astype('uint8')
# 保存分割结果
if save:
# 保存方式一: 直接保存\显示分割(预测)结果
if save_model == 1:
save_image_path = os.path.join(save_path, 'pred-' + img_path.split('\\')[-1])
cv2.imwrite(save_image_path, pred_mask_bgr)
# 保存方式二: 将分割(预测)结果和原图叠加保存
elif save_model == 2:
opacity = 0.7 # 透明度,取值范围(0,1),越大越接近原图
pred_add = cv2.addWeighted(img_bgr, opacity, pred_mask_bgr, 1-opacity, 0)
save_image_path = os.path.join(save_path, 'pred-' + img_path.split('/')[-1])
cv2.imwrite(save_image_path, pred_add)
# 保存方式三:检测分割结果的边界轮廓,然后叠加到原图上
elif save_model == 3:
# 预测图转为灰度图
binary = np.where(0.5 < pred_mask, 1, 0).astype(dtype=np.uint8)
binary = binary * 255
# 检测mask的边界
contours, hierarchy = cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# 绘制边界轮廓
# cv2.drawContours(image, contours, -1, (255, 0, 0), 1) # 蓝色 # (255, 0, 0) 蓝色； (0, 255, 0)绿色； (0, 0, 255)蓝色
cv2.drawContours(img_bgr, contours, -1, (0, 255, 0), 1) # 绿色
# 保存
# save_image_path = os.path.join(save_path, 'pred-' + img_path.split('\\')[-1])
save_image_path = os.path.join(save_path, img_path.split('\\')[-1])
cv2.imwrite(save_image_path, img_bgr)
# cv2.imencode(".png", img_bgr)[1].tofile(save_image_path)
# 显示分割结果
if show:
# #****** 可视化语义分割预测结果——方法一（直接显示分割结果）******#
if show_model == 1:
plt.figure(figsize=(8, 8))
plt.imshow(pred_mask)
plt.savefig('test_result/k1-0.jpg')
plt.show()
print(result.seg_logits.data.shape) # 定量
# #****** 可视化语义分割预测结果--方法二（叠加在原因上进行显示）******#
elif show_model == 2:
# 显示语义分割结果
plt.figure(figsize=(10, 8))
plt.imshow(img_bgr[:, :, ::-1])
plt.imshow(pred_mask, alpha=0.55)
plt.axis('off')
plt.savefig('test_result/k1-1.jpg')
plt.show()
# #****** 可视化语义分割预测结果--方法三(和原图并排显示) ******#
elif show_model == 3:
plt.figure(figsize=(14, 8))
plt.subplot(1, 2, 1)
plt.imshow(img_bgr[:, :, ::-1])
plt.axis('off')
plt.subplot(1, 2, 2)
plt.imshow(img_bgr[:, :, ::-1])
plt.imshow(pred_mask, alpha=0.6)
plt.axis('off')
plt.savefig('test_result/k1-2.jpg')
plt.show()
# #****** 可视化语义分割预测结果-方法四(按配色方案叠加在原图上显示) ******#
elif show_model == 4:
opacity = 0.3 # 透明度,越大越接近原图
# 将预测的整数ID,映射为对应类别的颜色
pred_mask_bgr = np.zeros((pred_mask.shape[0], pred_mask.shape[1], 3))
for idx in palette_dict.keys():
pred_mask_bgr[np.where(pred_mask == idx)] = palette_dict[idx]
pred_mask_bgr = pred_mask_bgr.astype('uint8')
# 将语义分割预测图和原图叠加显示
pred_viz = cv2.addWeighted(img_bgr, opacity, pred_mask_bgr, 1 - opacity, 0)
cv2.imwrite('test_result/k1-3.jpg', pred_viz)
plt.figure(figsize=(8, 8))
plt.imshow(pred_viz[:, :, ::-1])
plt.show()
# #***** 可视化语义分割预测结果-方法五(按mmseg/datasets/cag.py里定义的类别颜色可视化) ***** #
elif show_model == 5:
img_viz = show_result_pyplot(model, img_path, result, opacity=0.8, title='MMSeg',
out_file='test_result/k1-4.jpg')
print('the shape of img_viz:', img_viz.shape)
plt.figure(figsize=(14, 8))
plt.imshow(img_viz)
plt.show()
# #***** 可视化语义分割预测结果--方法六(加图例) ***** #
elif show_model == 6:
# 获取类别名和调色板
classes = CAGDataset.METAINFO['classes']
palette = CAGDataset.METAINFO['palette']
opacity = 0.15
# 将分割图按调色板染色
seg_map = pred_mask.astype('uint8')
seg_img = Image.fromarray(seg_map).convert('P')
seg_img.putpalette(np.array(palette, dtype=np.uint8))
# from matplotlib import pyplot as plt
# import matplotlib.patches as mpatches
plt.figure(figsize=(14, 8))
img_plot = ((np.array(seg_img.convert('RGB'))) * (1 - opacity) + mmcv.imread(img_path) * opacity) / 255
im = plt.imshow(img_plot)
# 为每一种颜色创建一个图例
patches = [mpatches.Patch(color=np.array(palette[i]) / 255, label=classes[i]) for i in range(len(classes))]
plt.legend(handles=patches, bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0., fontsize='large')
plt.savefig('test_result/k1-6.jpg')
plt.show()
if __name__ == '__main__':
# 指定测试集路径
test_images_path = '../data/XXX/images/test'
# 指定测试集结果存放路径
result_save_path = '../data/XXX/predict_result/Segformer_model3'
# 检测结果存放的目录是否存在.不存在的话,创建空文件夹
if not os.path.exists(result_save_path):
os.mkdir(result_save_path)
# 载入模型
# 模型config配置文件
config_file = '../my_Configs/my_Segformer_20230907.py'
# 模型checkpoint权重文件
checkpoint_file = '../work_dirs/my_Segformer/best_mIoU_iter_50000.pth'
# 计算硬件
device = 'cuda:0'
# 指定各个类别的BGR配色
palette = [
['background', [0, 0, 0]],
['vessel', [255, 255, 255]]
]
palette_dict = {}
for idx, each in enumerate(palette):
palette_dict[idx] = each[1]
print(palette_dict)
# 加载模型
model = init_model(config_file, checkpoint_file, device=device)
# 单张图像预测函数
opacity = 0.7
# 测试集批量预测
for each in tqdm(os.listdir(test_images_path)):
print(each)
image_path = os.path.join(test_images_path, each)
# predict_single_img(image_path, model, save=True, save_path=result_save_path, save_model=3, show=False)
predict_single_img(image_path, model, save=False, save_path=result_save_path, save_model=3, show=True, show_model=2)
扫一扫
1586
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
