pip install mmcv-full -f download.openmmlab/mmcv/dist/cu101/torch1.6.0/index.html
git clone github/open-mmlab/mmsegmentation.git
cd mmsegmentation
pip install -e . # or "python setup.py develop"
mkdir data
ln -s $DATA_ROOT data
由于⼯作需要,需要在MMSeg上训练⾃⼰的数据集,⽽之前只在上⾯训过Cityscapes的数据集,两种数据集质检的组织形式不同,需要⾃⼰重写配置⽂件,我的数据原始格式为⼀对img和label,两者⽂件名相同,后缀img.tif,label.png,如图: 原始图像:
标签:
⾸先在/mmsegmentation-master/mmseg/datasets路径下新建⼀个mydataset.py⽂件 @ister_module()
class mydata(CustomDataset):
CLASSES = ('background', 'plant', 'plantsoil', 'grass', 'building',
'railway', 'structure', 'humanland', 'baresoil', 'water')
PALETTE = [[0,0,0], [244, 35, 232], [70, 70, 70], [102, 102, 156],
[190, 153, 153], [153, 153, 153], [250, 170, 30], [220, 220, 0],
[107, 142, 35]]
def __init__(self, **kwargs):
super(mydata, self).__init__(
img_suffix='.tif',
seg_map_suffix='.png',
reduce_zero_label=False,
**kwargs)
ists(self.img_dir)
写法参考该⽂件夹下其他数据集的py⽂件,需要修改的地⽅为CLASSES和PALETTE,前者为你数据集中每⼀类的名称,后者为类别对应的颜⾊,注意不要漏掉背景值'background'。然后根据⾃⼰数据集的后缀更改img_suffix和seg_map_suffix,改好之后保存 同时还需要在同⼀⽂件夹下的__init__.py⽂件中添加刚刚新建的数据集
from .stare import STAREDataset
from .voc import PascalVOCDataset
from .mydataset import mydata
__all__ = [
光固化剂
'CustomDataset', 'Test_pathDataset','build_dataloader', 'ConcatDataset', 'RepeatDataset',
'DATASETS', 'build_dataset', 'PIPELINES', 'CityscapesDataset',
'PascalVOCDataset', 'ADE20KDataset', 'PascalContextDataset',
'PascalContextDataset59', 'ChaseDB1Dataset', 'DRIVEDataset', 'HRFDataset',
'STAREDataset', 'DarkZurichDataset', 'NightDrivingDataset',
'COCOStuffDataset', 'mydata'
炒茶机]
/mmsegmentation-master/mmseg/core/evaluation下的class_names.py⽂件也需要进⾏更改:import mmcv
def mydata_classes():
"""shengteng class names for external use."""
return [
'background', 'plant', 'plantsoil', 'grass', 'building',
'railway', 'structure', 'humanland', 'baresoil', 'water'
]
def mydata_palette():
return [[0,0,0], [244, 35, 232], [70, 70, 70], [102, 102, 156],
[190, 153, 153], [153, 153, 153], [250, 170, 30], [220, 220, 0],
[107, 142, 35]]
⾄此数据集更改完成,还需要更改config⽂件下的模型⽂件:
norm_cfg = dict(type='SyncBN', requires_grad=True)
backbone_norm_cfg = dict(type='LN', requires_grad=True)
model = dict(
type='EncoderDecoder',
pretrained='pretrain/swin_base_upernet_224x224_1K.pth', #从官⽹下载与训练模型 backbone=dict(
type='SwinTransformer',
pretrain_img_size=224,
embed_dims=128,
patch_size=4,
window_size=7,
mlp_ratio=4,
depths=[2, 2, 18, 2],
num_heads=[4, 8, 16, 32],
strides=(4, 2, 2, 2),阵列天线
out_indices=(0, 1, 2, 3),
qkv_bias=True,
qk_scale=None,
ca1703
patch_norm=True,
drop_rate=0.0,
attn_drop_rate=0.0,
drop_path_rate=0.3,
use_abs_pos_embed=False,
act_cfg=dict(type='GELU'),
norm_cfg=dict(type='LN', requires_grad=True)),
decode_head=dict(
type='UPerHead',
in_channels=[128, 256, 512, 1024],
in_index=[0, 1, 2, 3],
pool_scales=(1, 2, 3, 6),
channels=512,
高浓除砂器dropout_ratio=0.1,
num_classes=9, #安装⾃⼰数据集的类别数更改
norm_cfg=dict(type='SyncBN', requires_grad=True),
align_corners=False,
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
修改batch和数据集路径,train val test三部分都要更改:
data = dict(
samples_per_gpu=8, #batch size
workers_per_gpu=12, #每张GPU分配的CPU核⼼数
train=dict(
type='mydata', #⾃定义的数据集名称
data_root='data/xxx', #数据根⽬录
img_dir='img/', #原始图像路径
ann_dir='label/', #标签路径
pipeline=[
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=True),
dict(type='Resize', img_scale=(512, 512), ratio_range=(1.0, 1.0)),
dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(
type='Normalize',
流程工业在线mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg'])
])
修改完成后开始在⾃⼰的数据集上训练MMSeg:
python ./tools/train.py ./config/swin/upernet_swin_base_patch4_window7_512x512.py 训练的log和权重⽂件会保存在workdir⽂件下:
多GPU训练:
bash ./tools/dist_train.sh config.py 2 #config为你的配置⽂件,2为使⽤的GPU的数量