Detectron学习三：自定义数据层

# -*- coding: utf-8 -*-
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved

from fvcore.common.file_io import PathManager
import os
import numpy as np
import xml.etree.ElementTree as ET

from detectron2.structures import BoxMode
from detectron2.data import DatasetCatalog, MetadataCatalog


__all__ = ["register_pascal_voc"]


# fmt: off
CLASS_NAMES = [
    "aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat",
    "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person",
    "pottedplant", "sheep", "sofa", "train", "tvmonitor",
]   ##为xml中目标的name名称
# fmt: on


def load_voc_instances(dirname: str, split: str):
    """
    Load Pascal VOC detection annotations to Detectron2 format.

    Args:
        dirname: Contain "Annotations", "ImageSets", "JPEGImages"
        split (str): one of "train", "test", "val", "trainval"
    """
    with PathManager.open(os.path.join(dirname, "ImageSets", "Main", split + ".txt")) as f:
        fileids = np.loadtxt(f, dtype=np.str)

    dicts = []
    ######################
    xml解析（省略）
    ######################
    return dicts

def register_pascal_voc(name, dirname, split, year):
    DatasetCatalog.register(name, lambda: load_voc_instances(dirname, split))
    MetadataCatalog.get(name).set(
        thing_classes=CLASS_NAMES, dirname=dirname, year=year, split=split
    )

# ==== Predefined splits for PASCAL VOC ===========
def register_all_pascal_voc(root="datasets"):
    SPLITS = [
        ("voc_2007_trainval", "VOC2007", "trainval"),
        ("voc_2007_train", "VOC2007", "train"),
        ("voc_2007_val", "VOC2007", "val"),
        ("voc_2007_test", "VOC2007", "test"),
        ("voc_2012_trainval", "VOC2012", "trainval"),
        ("voc_2012_train", "VOC2012", "train"),
        ("voc_2012_val", "VOC2012", "val"),
    ]   ##该列表中的名称与configs中yaml中DATASETS.{TRAIN,TEST}中的名字完全一致
    
    for name, dirname, split in SPLITS:
        year = 2007 if "2007" in name else 2012
        register_pascal_voc(name, os.path.join(root, dirname), split, year)
        MetadataCatalog.get(name).evaluator_type = "pascal_voc"

标准数据格式：datasetFormat=[dict1,dict2,dict3,....]  ##len(datasetFormat)为数据集大小，一张图像对应一个dict
dict1={
    "file_name":str, ##图像绝对路径
    "image_id":str or int,
    "annotations":[ ##list[dict]
        {
            "category_id":int,  ##类别索引，CLASS_NAMES.index(cls)的返回值，cls为类别名name
            "bbox":[float,float,float,float], ##排列顺序与下面bbox_mode有关
            "bbox_mode":int    ##XYXY_ABS:[xmin,ymin,xmax,ymax]  XYWH_ABS:[xmin,ymin,width,height]
                                ##XYXY_REL XYWH 相对与图像尺寸的坐标值，范围[0,1]
        },
        {
        }，
    ]    
}

def register_pascal_voc(name, dirname, split, year):
    DatasetCatalog.register(name, lambda: load_voc_instances(dirname, split))
 
 #################################
 类DatasetCatalog中register函数的定义
 @staticmethod
 def register(name,func):
        """
        Args:
            name (str): the name that identifies a dataset, e.g. "coco_2014_train".
            func (callable): a callable which takes no arguments and returns a list of dicts.
        """
  #############################
  此处使用了lambda表达式
  作用：借助lambda给回调函数传参，做了一个“函数打包”的操作，它把函数和它的参数列表打包在一起，得到一个新的函数，
  实际传入到register的并不是原来的load_voc_instances，而是打包得到的新函数。至于这个新函数叫做什么名字，对于
  我们来说是无所谓的（匿名函数），反正也只用这一次，下次还要用的话，lambda会再一次重新打包，得到新的匿名函数。

from detectron2.data import MetadataCatalog
MetadataCatalog.get("my_dataset").thing_classes = ["person", "dog"]

# 也可以这样
MetadataCatalog.get("my_dataset").set("thing_classes",["person", "dog"])