（六）mmdetection源码解读：datasets = [build_dataset(cfg.data.train)]

一、build_dataset

build_dataset函数定义

def build_dataset(cfg, default_args=None): 
    dataset = build_from_cfg(cfg, DATASETS, default_args)
    return dataset

二、build_from_cfg

build_from_cfg函数中实现了使用配置参数实例化数据集类

def build_from_cfg(cfg, registry, default_args=None):
      return obj_cls(**args

三、class KittiTinyDataset(CustomDataset)

数据集定义，主要的实例化都是在父类CustomDataset中实现的

@DATASETS.register_module()
class KittiTinyDataset(CustomDataset):
    CLASSES = ('Car', 'Pedestrian', 'Cyclist')
    def load_annotations(self, ann_file):
    	...

四、class CustomDataset(Dataset)

CustomDataset会根据配置文件初始化图像和标注文件相关变量，加载标注信息load_annotations，最重要的部分是实例化self.pipeline，也就是数据的处理流程

@DATASETS.register_module()
class CustomDataset(Dataset):
  
    CLASSES = None

    def __init__(self,
                 ann_file,
                 pipeline,
                 classes=None,
                 data_root=None,
                 img_prefix='',
                 seg_prefix=None,
                 proposal_file=None,
                 test_mode=False,
                 filter_empty_gt=True,
                 file_client_args=dict(backend='disk')):
        self.ann_file = ann_file
        self.data_root = data_root
        self.img_prefix = img_prefix
        self.seg_prefix = seg_prefix
        self.proposal_file = proposal_file
        self.test_mode = test_mode
        self.filter_empty_gt = filter_empty_gt
        self.CLASSES = self.get_classes(classes)
        self.file_client = mmcv.FileClient(**file_client_args)


        if hasattr(self.file_client, 'get_local_path'):
            with self.file_client.get_local_path(self.ann_file) as local_path:
                self.data_infos = self.load_annotations(local_path)
       

        # processing pipeline
        self.pipeline = Compose(pipeline)

五、Compose

因为pipeline是一个list，我们会根据list的内容逐个 去实例化pipeline中的类LoadImageFromFile、LoadAnnotations、Resize、RandomFlip、Normalize、Pad、DefaultFormatBundle、Collect等

@PIPELINES.register_module()
class Compose:
    """Compose multiple transforms sequentially.

    Args:
        transforms (Sequence[dict | callable]): Sequence of transform object or
            config dict to be composed.
    """

    def __init__(self, transforms):
        assert isinstance(transforms, collections.abc.Sequence)
        self.transforms = []
        for transform in transforms:
            if isinstance(transform, dict):
                transform = build_from_cfg(transform, PIPELINES)
                self.transforms.append(transform)
            elif callable(transform):
                self.transforms.append(transform)
            else:
                raise TypeError('transform must be callable or a dict')

六、build_from_cfg(transform, PIPELINES)

最后再调用build_from_cfg实例化PIPELINES的所有流程