Config

import json
from mmcv import Config


cfg_path='./config/pan_r18_ctw.py'

cfg=Config.fromfile(cfg_path)

#读取参数，json.dumps()使字典类型漂亮的输出，indent参数决定添加几个空格
print(json.dumps(cfg._cfg_dict,indent=4))

{
    "model": {
        "type": "PAN",
        "backbone": {
            "type": "resnet18",
            "pretrained": true
        },
        "neck": {
            "type": "FPEM_v1",
            "in_channels": [
                64,
                128,
                256,
                512
            ],
            "out_channels": 128
        },
        "detection_head": {
            "type": "PA_Head",
            "in_channels": 512,
            "hidden_dim": 128,
            "num_classes": 6,
            "loss_text": {
                "type": "DiceLoss",
                "loss_weight": 1.0
            },
            "loss_kernel": {
                "type": "DiceLoss",
                "loss_weight": 0.5
            },
            "loss_emb": {
                "type": "EmbLoss_v1",
                "feature_dim": 4,
                "loss_weight": 0.25
            }
        }
    },
    "data": {
        "batch_size": 16,
        "train": {
            "type": "PAN_CTW",
            "split": "train",
            "is_transform": true,
            "img_size": 640,
            "short_size": 640,
            "kernel_scale": 0.7,
            "read_type": "cv2"
        },
        "test": {
            "type": "PAN_CTW",
            "split": "test",
            "short_size": 640,
            "read_type": "cv2"
        }
    },
    "train_cfg": {
        "lr": 0.001,
        "schedule": "polylr",
        "epoch": 600,
        "optimizer": "Adam"
    },
    "test_cfg": {
        "min_score": 0.88,
        "min_area": 16,
        "bbox_type": "poly",
        "result_path": "outputs/submit_ctw/"
    }
}

选取某一个模块

print(cfg.data.train)

{'type': 'PAN_CTW', 'split': 'train', 'is_transform': True, 'img_size': 640, 'short_size': 640, 'kernel_scale': 0.7, 'read_type': 'cv2'}

type(cfg.data.train)

<class 'mmcv.utils.config.ConfigDict'>

导入一个模块

1. 建立builder.py文件

import models

def build_model(cfg):
        param=dict()
        for key in cfg:
            if key=='type':
                continue
            param[key]=cfg[key]
        model=models.__dict__[cfg.type](**param)
        return model

当需要使用config中的backbone参数信息时，还需在builder.py所在位init.py修改

from .builder import budil_backbone
from .resnet import resnet18,resnet50,resnet101

__all__=['resnet18','resnet50','resnet101','budil_backbone']
#定义__all__变量，默认对外允许导入以下四个函数

budil_backbone(cfg.model.backbone)

该函数的解析如下

import models

def budil_backbone(cfg):

    param=dict()
    for key in cfg:
        if key=='type':
            continue
        param[key]=cfg[key]

    #print(cfg.type)
    #print(*param) pretrained
    #{'pretrained': True}
    #print(models.backbone.__dict__[cfg.type]) 此处对应的时dict['resnet18']
    #<function resnet18 at 0x000001C7A55A79D0>
    #print(models.backbone.__dict__)
    # {'type': 'PAN_CTW', 'split': 'train', 'is_transform': True, 'img_size': 640, 'short_size': 640, 'kernel_scale': 0.7, 'read_type': 'cv2'}
    #backbone=models.backbone.__dict__[cfg.type](param.values())
    backbone = models.backbone.__dict__[cfg.type](**param)
    #functional 传入parma的value值

    return backbone