compressai_vision.datasets#

class compressai_vision.datasets.DataCatalog(root, imgs_folder='images', annotation_file='sample.json', seqinfo='seqinfo.ini', dataset_name='sample_dataset', ext='.png')[source]#
property annotation_path#
property dataset#
property dataset_name#
property imgs_folder_path#
property seqinfo_path#
class compressai_vision.datasets.DefaultDataset(root, dataset_name, imgs_folder: str = 'valid', **kwargs)[source]#

Loads an image folder database. testing image samples are respectively stored in separate directories (Currently, this class does not support any of the training related operations):

|--rootdir
  |-- img000.png
  |-- img001.png
Variables:

  • root (string) – root directory of the dataset

  • transform ((callable, optional)) – a function or transform that takes in a PIL image and returns a transformed version

  • use_BGR (Bool) – if True the color order of the sample is BGR otherwise RGB returned

class compressai_vision.datasets.Detectron2Dataset(root, dataset_name, imgs_folder, **kwargs)[source]#
get_org_mapper_func()[source]#
class compressai_vision.datasets.TrackingDataset(root, dataset_name, imgs_folder, **kwargs)[source]#
get_org_mapper_func()[source]#
compressai_vision.datasets.deccode_compressed_rle(data)[source]#
compressai_vision.datasets.get_seq_info(seq_info_path)[source]#