text
stringlengths
81
112k
Reads and decodes an image from a file object as a Numpy array. The SUN dataset contains images in several formats (despite the fact that all of them have .jpg extension). Some of them are: - BMP (RGB) - PNG (grayscale, RGBA, RGB interlaced) - JPEG (RGB) - GIF (1-frame RGB) Since TFDS assumes that all images have the same number of channels, we convert all of them to RGB. Args: fobj: File object to read from. session: TF session used to decode the images. filename: Filename of the original image in the archive. Returns: Numpy array with shape (height, width, channels). def _decode_image(fobj, session, filename): """Reads and decodes an image from a file object as a Numpy array. The SUN dataset contains images in several formats (despite the fact that all of them have .jpg extension). Some of them are: - BMP (RGB) - PNG (grayscale, RGBA, RGB interlaced) - JPEG (RGB) - GIF (1-frame RGB) Since TFDS assumes that all images have the same number of channels, we convert all of them to RGB. Args: fobj: File object to read from. session: TF session used to decode the images. filename: Filename of the original image in the archive. Returns: Numpy array with shape (height, width, channels). """ buf = fobj.read() image = tfds.core.lazy_imports.cv2.imdecode( np.fromstring(buf, dtype=np.uint8), flags=3) # Note: Converts to RGB. if image is None: logging.warning( "Image %s could not be decoded by OpenCV, falling back to TF", filename) try: image = tf.image.decode_image(buf, channels=3) image = session.run(image) except tf.errors.InvalidArgumentError: logging.fatal("Image %s could not be decoded by Tensorflow", filename) # The GIF images contain a single frame. if len(image.shape) == 4: # rank=4 -> rank=3 image = image.reshape(image.shape[1:]) return image
Process image files from the dataset. def _process_image_file(fobj, session, filename): """Process image files from the dataset.""" # We need to read the image files and convert them to JPEG, since some files # actually contain GIF, PNG or BMP data (despite having a .jpg extension) and # some encoding options that will make TF crash in general. image = _decode_image(fobj, session, filename=filename) return _encode_jpeg(image)
Yields examples. def _generate_examples(self, archive): """Yields examples.""" prefix_len = len("SUN397") with tf.Graph().as_default(): with utils.nogpu_session() as sess: for filepath, fobj in archive: if (filepath.endswith(".jpg") and filepath not in _SUN397_IGNORE_IMAGES): # Note: all files in the tar.gz are in SUN397/... filename = filepath[prefix_len:] # Example: # From filename: /c/car_interior/backseat/sun_aenygxwhhmjtisnf.jpg # To class: /c/car_interior/backseat label = "/".join(filename.split("/")[:-1]) image = _process_image_file(fobj, sess, filepath) yield { "file_name": filename, "image": image, "label": label, }
Returns examples from parallel SGML or text files, which may be gzipped. def _parse_parallel_sentences(f1, f2): """Returns examples from parallel SGML or text files, which may be gzipped.""" def _parse_text(path): """Returns the sentences from a single text file, which may be gzipped.""" split_path = path.split(".") if split_path[-1] == "gz": lang = split_path[-2] with tf.io.gfile.GFile(path) as f, gzip.GzipFile(fileobj=f) as g: return g.read().split("\n"), lang if split_path[-1] == "txt": # CWMT lang = split_path[-2].split("_")[-1] lang = "zh" if lang in ("ch", "cn") else lang else: lang = split_path[-1] with tf.io.gfile.GFile(path) as f: return f.read().split("\n"), lang def _parse_sgm(path): """Returns sentences from a single SGML file.""" lang = path.split(".")[-2] sentences = [] # Note: We can't use the XML parser since some of the files are badly # formatted. seg_re = re.compile(r"<seg id=\"\d+\">(.*)</seg>") with tf.io.gfile.GFile(path) as f: for line in f: seg_match = re.match(seg_re, line) if seg_match: assert len(seg_match.groups()) == 1 sentences.append(seg_match.groups()[0]) return sentences, lang parse_file = _parse_sgm if f1.endswith(".sgm") else _parse_text # Some datasets (e.g., CWMT) contain multiple parallel files specified with # a wildcard. We sort both sets to align them and parse them one by one. f1_files = tf.io.gfile.glob(f1) f2_files = tf.io.gfile.glob(f2) assert f1_files and f2_files, "No matching files found: %s, %s." % (f1, f2) assert len(f1_files) == len(f2_files), ( "Number of files do not match: %d vs %d for %s vs %s." % ( len(f1_files), len(f2_files), f1, f2)) for f1_i, f2_i in zip(sorted(f1_files), sorted(f2_files)): l1_sentences, l1 = parse_file(f1_i) l2_sentences, l2 = parse_file(f2_i) assert len(l1_sentences) == len(l2_sentences), ( "Sizes do not match: %d vs %d for %s vs %s." % ( len(l1_sentences), len(l2_sentences), f1_i, f2_i)) for s1, s2 in zip(l1_sentences, l2_sentences): yield { l1: s1, l2: s2 }
Generates examples from TMX file. def _parse_tmx(path): """Generates examples from TMX file.""" def _get_tuv_lang(tuv): for k, v in tuv.items(): if k.endswith("}lang"): return v raise AssertionError("Language not found in `tuv` attributes.") def _get_tuv_seg(tuv): segs = tuv.findall("seg") assert len(segs) == 1, "Invalid number of segments: %d" % len(segs) return segs[0].text with tf.io.gfile.GFile(path) as f: for _, elem in ElementTree.iterparse(f): if elem.tag == "tu": yield { _get_tuv_lang(tuv): _get_tuv_seg(tuv) for tuv in elem.iterfind("tuv") } elem.clear()
Generates examples from TSV file. def _parse_tsv(path, language_pair=None): """Generates examples from TSV file.""" if language_pair is None: lang_match = re.match(r".*\.([a-z][a-z])-([a-z][a-z])\.tsv", path) assert lang_match is not None, "Invalid TSV filename: %s" % path l1, l2 = lang_match.groups() else: l1, l2 = language_pair with tf.io.gfile.GFile(path) as f: for j, line in enumerate(f): cols = line.split("\t") if len(cols) != 2: logging.warning( "Skipping line %d in TSV (%s) with %d != 2 columns.", j, path, len(cols)) continue s1, s2 = cols yield { l1: s1.strip(), l2: s2.strip() }
Generates examples from Wikiheadlines dataset file. def _parse_wikiheadlines(path): """Generates examples from Wikiheadlines dataset file.""" lang_match = re.match(r".*\.([a-z][a-z])-([a-z][a-z])$", path) assert lang_match is not None, "Invalid Wikiheadlines filename: %s" % path l1, l2 = lang_match.groups() with tf.io.gfile.GFile(path) as f: for line in f: s1, s2 = line.split("|||") yield { l1: s1.strip(), l2: s2.strip() }
Generates examples from CzEng v1.6, with optional filtering for v1.7. def _parse_czeng(*paths, **kwargs): """Generates examples from CzEng v1.6, with optional filtering for v1.7.""" filter_path = kwargs.get("filter_path", None) if filter_path: re_block = re.compile(r"^[^-]+-b(\d+)-\d\d[tde]") with tf.io.gfile.GFile(filter_path) as f: bad_blocks = { blk for blk in re.search( r"qw{([\s\d]*)}", f.read()).groups()[0].split() } logging.info( "Loaded %d bad blocks to filter from CzEng v1.6 to make v1.7.", len(bad_blocks)) for path in paths: for gz_path in tf.io.gfile.glob(path): with tf.io.gfile.GFile(gz_path, "rb") as g, gzip.GzipFile(fileobj=g) as f: for line in f: line = line.decode("utf-8") # required for py3 if not line.strip(): continue id_, unused_score, cs, en = line.split("\t") if filter_path: block_match = re.match(re_block, id_) if block_match and block_match.groups()[0] in bad_blocks: continue yield { "cs": cs.strip(), "en": en.strip(), }
Injects languages into (potentially) template strings. def _inject_language(self, src, strings): """Injects languages into (potentially) template strings.""" if src not in self.sources: raise ValueError("Invalid source for '{0}': {1}".format(self.name, src)) def _format_string(s): if "{0}" in s and "{1}" and "{src}" in s: return s.format(*sorted([src, self.target]), src=src) elif "{0}" in s and "{1}" in s: return s.format(*sorted([src, self.target])) elif "{src}" in s: return s.format(src=src) else: return s return [_format_string(s) for s in strings]
Subsets that make up each split of the dataset for the language pair. def subsets(self): """Subsets that make up each split of the dataset for the language pair.""" source, target = self.builder_config.language_pair filtered_subsets = {} for split, ss_names in self._subsets.items(): filtered_subsets[split] = [] for ss_name in ss_names: ds = DATASET_MAP[ss_name] if ds.target != target or source not in ds.sources: logging.info( "Skipping sub-dataset that does not include language pair: %s", ss_name) else: filtered_subsets[split].append(ss_name) logging.info("Using sub-datasets: %s", filtered_subsets) return filtered_subsets
Returns the examples in the raw (text) form. def _generate_examples(self, split_subsets, extraction_map): """Returns the examples in the raw (text) form.""" source, _ = self.builder_config.language_pair def _get_local_paths(ds, extract_dirs): rel_paths = ds.get_path(source) if len(extract_dirs) == 1: extract_dirs = extract_dirs * len(rel_paths) return [os.path.join(ex_dir, rel_path) if rel_path else ex_dir for ex_dir, rel_path in zip(extract_dirs, rel_paths)] for ss_name in split_subsets: logging.info("Generating examples from: %s", ss_name) ds = DATASET_MAP[ss_name] extract_dirs = extraction_map[ss_name] files = _get_local_paths(ds, extract_dirs) if ss_name.startswith("czeng"): if ss_name.endswith("16pre"): sub_generator = functools.partial( _parse_tsv, language_pair=("en", "cs")) elif ss_name.endswith("17"): filter_path = _get_local_paths( _CZENG17_FILTER, extraction_map[_CZENG17_FILTER.name])[0] sub_generator = functools.partial( _parse_czeng, filter_path=filter_path) else: sub_generator = _parse_czeng elif len(files) == 2: if ss_name.endswith("_frde"): sub_generator = _parse_frde_bitext else: sub_generator = _parse_parallel_sentences elif len(files) == 1: fname = files[0] # Note: Due to formatting used by `download_manager`, the file # extension may not be at the end of the file path. if ".tsv" in fname: sub_generator = _parse_tsv elif ss_name.startswith("newscommentary_v14"): sub_generator = functools.partial( _parse_tsv, language_pair=self.builder_config.language_pair) elif "tmx" in fname: sub_generator = _parse_tmx elif ss_name.startswith("wikiheadlines"): sub_generator = _parse_wikiheadlines else: raise ValueError("Unsupported file format: %s" % fname) else: raise ValueError("Invalid number of files: %d" % len(files)) for ex in sub_generator(*files): if not all(ex.values()): continue # TODO(adarob): Add subset feature. # ex["subset"] = subset yield ex
Fetches a `tfds.core.DatasetBuilder` by string name. Args: name: `str`, the registered name of the `DatasetBuilder` (the snake case version of the class name). This can be either `"dataset_name"` or `"dataset_name/config_name"` for datasets with `BuilderConfig`s. As a convenience, this string may contain comma-separated keyword arguments for the builder. For example `"foo_bar/a=True,b=3"` would use the `FooBar` dataset passing the keyword arguments `a=True` and `b=3` (for builders with configs, it would be `"foo_bar/zoo/a=True,b=3"` to use the `"zoo"` config and pass to the builder keyword arguments `a=True` and `b=3`). **builder_init_kwargs: `dict` of keyword arguments passed to the `DatasetBuilder`. These will override keyword arguments passed in `name`, if any. Returns: A `tfds.core.DatasetBuilder`. Raises: DatasetNotFoundError: if `name` is unrecognized. def builder(name, **builder_init_kwargs): """Fetches a `tfds.core.DatasetBuilder` by string name. Args: name: `str`, the registered name of the `DatasetBuilder` (the snake case version of the class name). This can be either `"dataset_name"` or `"dataset_name/config_name"` for datasets with `BuilderConfig`s. As a convenience, this string may contain comma-separated keyword arguments for the builder. For example `"foo_bar/a=True,b=3"` would use the `FooBar` dataset passing the keyword arguments `a=True` and `b=3` (for builders with configs, it would be `"foo_bar/zoo/a=True,b=3"` to use the `"zoo"` config and pass to the builder keyword arguments `a=True` and `b=3`). **builder_init_kwargs: `dict` of keyword arguments passed to the `DatasetBuilder`. These will override keyword arguments passed in `name`, if any. Returns: A `tfds.core.DatasetBuilder`. Raises: DatasetNotFoundError: if `name` is unrecognized. """ name, builder_kwargs = _dataset_name_and_kwargs_from_name_str(name) builder_kwargs.update(builder_init_kwargs) if name in _ABSTRACT_DATASET_REGISTRY: raise DatasetNotFoundError(name, is_abstract=True) if name in _IN_DEVELOPMENT_REGISTRY: raise DatasetNotFoundError(name, in_development=True) if name not in _DATASET_REGISTRY: raise DatasetNotFoundError(name) try: return _DATASET_REGISTRY[name](**builder_kwargs) except BaseException: logging.error("Failed to construct dataset %s", name) raise
Loads the named dataset into a `tf.data.Dataset`. If `split=None` (the default), returns all splits for the dataset. Otherwise, returns the specified split. `load` is a convenience method that fetches the `tfds.core.DatasetBuilder` by string name, optionally calls `DatasetBuilder.download_and_prepare` (if `download=True`), and then calls `DatasetBuilder.as_dataset`. This is roughly equivalent to: ``` builder = tfds.builder(name, data_dir=data_dir, **builder_kwargs) if download: builder.download_and_prepare(**download_and_prepare_kwargs) ds = builder.as_dataset( split=split, as_supervised=as_supervised, **as_dataset_kwargs) if with_info: return ds, builder.info return ds ``` If you'd like NumPy arrays instead of `tf.data.Dataset`s or `tf.Tensor`s, you can pass the return value to `tfds.as_numpy`. Callers must pass arguments as keyword arguments. **Warning**: calling this function might potentially trigger the download of hundreds of GiB to disk. Refer to the `download` argument. Args: name: `str`, the registered name of the `DatasetBuilder` (the snake case version of the class name). This can be either `"dataset_name"` or `"dataset_name/config_name"` for datasets with `BuilderConfig`s. As a convenience, this string may contain comma-separated keyword arguments for the builder. For example `"foo_bar/a=True,b=3"` would use the `FooBar` dataset passing the keyword arguments `a=True` and `b=3` (for builders with configs, it would be `"foo_bar/zoo/a=True,b=3"` to use the `"zoo"` config and pass to the builder keyword arguments `a=True` and `b=3`). split: `tfds.Split` or `str`, which split of the data to load. If None, will return a `dict` with all splits (typically `tfds.Split.TRAIN` and `tfds.Split.TEST`). data_dir: `str` (optional), directory to read/write data. Defaults to "~/tensorflow_datasets". batch_size: `int`, set to > 1 to get batches of examples. Note that variable length features will be 0-padded. If `batch_size=-1`, will return the full dataset as `tf.Tensor`s. download: `bool` (optional), whether to call `tfds.core.DatasetBuilder.download_and_prepare` before calling `tf.DatasetBuilder.as_dataset`. If `False`, data is expected to be in `data_dir`. If `True` and the data is already in `data_dir`, `download_and_prepare` is a no-op. as_supervised: `bool`, if `True`, the returned `tf.data.Dataset` will have a 2-tuple structure `(input, label)` according to `builder.info.supervised_keys`. If `False`, the default, the returned `tf.data.Dataset` will have a dictionary with all the features. with_info: `bool`, if True, tfds.load will return the tuple (tf.data.Dataset, tfds.core.DatasetInfo) containing the info associated with the builder. builder_kwargs: `dict` (optional), keyword arguments to be passed to the `tfds.core.DatasetBuilder` constructor. `data_dir` will be passed through by default. download_and_prepare_kwargs: `dict` (optional) keyword arguments passed to `tfds.core.DatasetBuilder.download_and_prepare` if `download=True`. Allow to control where to download and extract the cached data. If not set, cache_dir and manual_dir will automatically be deduced from data_dir. as_dataset_kwargs: `dict` (optional), keyword arguments passed to `tfds.core.DatasetBuilder.as_dataset`. `split` will be passed through by default. Example: `{'shuffle_files': True}`. Note that shuffle_files is False by default unless `split == tfds.Split.TRAIN`. try_gcs: `bool`, if True, tfds.load will see if the dataset exists on the public GCS bucket before building it locally. Returns: ds: `tf.data.Dataset`, the dataset requested, or if `split` is None, a `dict<key: tfds.Split, value: tfds.data.Dataset>`. If `batch_size=-1`, these will be full datasets as `tf.Tensor`s. ds_info: `tfds.core.DatasetInfo`, if `with_info` is True, then `tfds.load` will return a tuple `(ds, ds_info)` containing dataset information (version, features, splits, num_examples,...). Note that the `ds_info` object documents the entire dataset, regardless of the `split` requested. Split-specific information is available in `ds_info.splits`. def load(name, split=None, data_dir=None, batch_size=1, download=True, as_supervised=False, with_info=False, builder_kwargs=None, download_and_prepare_kwargs=None, as_dataset_kwargs=None, try_gcs=False): """Loads the named dataset into a `tf.data.Dataset`. If `split=None` (the default), returns all splits for the dataset. Otherwise, returns the specified split. `load` is a convenience method that fetches the `tfds.core.DatasetBuilder` by string name, optionally calls `DatasetBuilder.download_and_prepare` (if `download=True`), and then calls `DatasetBuilder.as_dataset`. This is roughly equivalent to: ``` builder = tfds.builder(name, data_dir=data_dir, **builder_kwargs) if download: builder.download_and_prepare(**download_and_prepare_kwargs) ds = builder.as_dataset( split=split, as_supervised=as_supervised, **as_dataset_kwargs) if with_info: return ds, builder.info return ds ``` If you'd like NumPy arrays instead of `tf.data.Dataset`s or `tf.Tensor`s, you can pass the return value to `tfds.as_numpy`. Callers must pass arguments as keyword arguments. **Warning**: calling this function might potentially trigger the download of hundreds of GiB to disk. Refer to the `download` argument. Args: name: `str`, the registered name of the `DatasetBuilder` (the snake case version of the class name). This can be either `"dataset_name"` or `"dataset_name/config_name"` for datasets with `BuilderConfig`s. As a convenience, this string may contain comma-separated keyword arguments for the builder. For example `"foo_bar/a=True,b=3"` would use the `FooBar` dataset passing the keyword arguments `a=True` and `b=3` (for builders with configs, it would be `"foo_bar/zoo/a=True,b=3"` to use the `"zoo"` config and pass to the builder keyword arguments `a=True` and `b=3`). split: `tfds.Split` or `str`, which split of the data to load. If None, will return a `dict` with all splits (typically `tfds.Split.TRAIN` and `tfds.Split.TEST`). data_dir: `str` (optional), directory to read/write data. Defaults to "~/tensorflow_datasets". batch_size: `int`, set to > 1 to get batches of examples. Note that variable length features will be 0-padded. If `batch_size=-1`, will return the full dataset as `tf.Tensor`s. download: `bool` (optional), whether to call `tfds.core.DatasetBuilder.download_and_prepare` before calling `tf.DatasetBuilder.as_dataset`. If `False`, data is expected to be in `data_dir`. If `True` and the data is already in `data_dir`, `download_and_prepare` is a no-op. as_supervised: `bool`, if `True`, the returned `tf.data.Dataset` will have a 2-tuple structure `(input, label)` according to `builder.info.supervised_keys`. If `False`, the default, the returned `tf.data.Dataset` will have a dictionary with all the features. with_info: `bool`, if True, tfds.load will return the tuple (tf.data.Dataset, tfds.core.DatasetInfo) containing the info associated with the builder. builder_kwargs: `dict` (optional), keyword arguments to be passed to the `tfds.core.DatasetBuilder` constructor. `data_dir` will be passed through by default. download_and_prepare_kwargs: `dict` (optional) keyword arguments passed to `tfds.core.DatasetBuilder.download_and_prepare` if `download=True`. Allow to control where to download and extract the cached data. If not set, cache_dir and manual_dir will automatically be deduced from data_dir. as_dataset_kwargs: `dict` (optional), keyword arguments passed to `tfds.core.DatasetBuilder.as_dataset`. `split` will be passed through by default. Example: `{'shuffle_files': True}`. Note that shuffle_files is False by default unless `split == tfds.Split.TRAIN`. try_gcs: `bool`, if True, tfds.load will see if the dataset exists on the public GCS bucket before building it locally. Returns: ds: `tf.data.Dataset`, the dataset requested, or if `split` is None, a `dict<key: tfds.Split, value: tfds.data.Dataset>`. If `batch_size=-1`, these will be full datasets as `tf.Tensor`s. ds_info: `tfds.core.DatasetInfo`, if `with_info` is True, then `tfds.load` will return a tuple `(ds, ds_info)` containing dataset information (version, features, splits, num_examples,...). Note that the `ds_info` object documents the entire dataset, regardless of the `split` requested. Split-specific information is available in `ds_info.splits`. """ name, name_builder_kwargs = _dataset_name_and_kwargs_from_name_str(name) name_builder_kwargs.update(builder_kwargs or {}) builder_kwargs = name_builder_kwargs # Set data_dir if try_gcs and gcs_utils.is_dataset_on_gcs(name): data_dir = constants.GCS_DATA_DIR elif data_dir is None: data_dir = constants.DATA_DIR dbuilder = builder(name, data_dir=data_dir, **builder_kwargs) if download: download_and_prepare_kwargs = download_and_prepare_kwargs or {} dbuilder.download_and_prepare(**download_and_prepare_kwargs) if as_dataset_kwargs is None: as_dataset_kwargs = {} as_dataset_kwargs = dict(as_dataset_kwargs) as_dataset_kwargs["split"] = split as_dataset_kwargs["as_supervised"] = as_supervised as_dataset_kwargs["batch_size"] = batch_size ds = dbuilder.as_dataset(**as_dataset_kwargs) if with_info: return ds, dbuilder.info return ds
Extract kwargs from name str. def _dataset_name_and_kwargs_from_name_str(name_str): """Extract kwargs from name str.""" res = _NAME_REG.match(name_str) if not res: raise ValueError(_NAME_STR_ERR.format(name_str)) name = res.group("dataset_name") kwargs = _kwargs_str_to_kwargs(res.group("kwargs")) try: for attr in ["config", "version"]: val = res.group(attr) if val is None: continue if attr in kwargs: raise ValueError("Dataset %s: cannot pass %s twice." % (name, attr)) kwargs[attr] = val return name, kwargs except: logging.error(_NAME_STR_ERR.format(name_str)) # pylint: disable=logging-format-interpolation raise
Try cast to int, float, bool, str, in that order. def _cast_to_pod(val): """Try cast to int, float, bool, str, in that order.""" bools = {"True": True, "False": False} if val in bools: return bools[val] try: return int(val) except ValueError: try: return float(val) except ValueError: return tf.compat.as_text(val)
Try importing a module, with an informative error message on failure. def _try_import(module_name): """Try importing a module, with an informative error message on failure.""" try: mod = importlib.import_module(module_name) return mod except ImportError: err_msg = ("Tried importing %s but failed. See setup.py extras_require. " "The dataset you are trying to use may have additional " "dependencies.") utils.reraise(err_msg)
Returns list from list, tuple or ndarray. def np_to_list(elem): """Returns list from list, tuple or ndarray.""" if isinstance(elem, list): return elem elif isinstance(elem, tuple): return list(elem) elif isinstance(elem, np.ndarray): return list(elem) else: raise ValueError( 'Input elements of a sequence should be either a numpy array, a ' 'python list or tuple. Got {}'.format(type(elem)))
Transpose a nested dict[list] into a list[nested dict]. def _transpose_dict_list(dict_list): """Transpose a nested dict[list] into a list[nested dict].""" # 1. Unstack numpy arrays into list dict_list = utils.map_nested(np_to_list, dict_list, dict_only=True) # 2. Extract the sequence length (and ensure the length is constant for all # elements) length = {'value': None} # dict because `nonlocal` is Python3 only def update_length(elem): if length['value'] is None: length['value'] = len(elem) elif length['value'] != len(elem): raise ValueError( 'The length of all elements of one sequence should be the same. ' 'Got {} != {}'.format(length['value'], len(elem))) return elem utils.map_nested(update_length, dict_list, dict_only=True) # 3. Extract each individual elements return [ utils.map_nested(lambda elem: elem[i], dict_list, dict_only=True) # pylint: disable=cell-var-from-loop for i in range(length['value']) ]
See base class for details. def get_tensor_info(self): """See base class for details.""" # Add the additional length dimension to every shape def add_length_dim(tensor_info): return feature_lib.TensorInfo( shape=(self._length,) + tensor_info.shape, dtype=tensor_info.dtype, ) tensor_info = super(SequenceDict, self).get_tensor_info() return utils.map_nested(add_length_dim, tensor_info)
See base class for details. def get_serialized_info(self): """See base class for details.""" # Add the additional length dimension to every serialized features def add_length_dim(serialized_info): """Add the length dimension to the serialized_info. Args: serialized_info: One of tf.io.FixedLenFeature, tf.io.VarLenFeature,... Returns: new_serialized_info: serialized_info with extended first dimension """ if isinstance(serialized_info, tf.io.FixedLenFeature): if self._length is not None: return tf.io.FixedLenFeature( shape=(self._length,) + serialized_info.shape, dtype=serialized_info.dtype, ) else: return tf.io.FixedLenSequenceFeature( shape=serialized_info.shape, dtype=serialized_info.dtype, allow_missing=True, ) elif isinstance(serialized_info, tf.io.VarLenFeature): return serialized_info else: raise ValueError( 'FixedLenSequenceFeature not supported inside SequenceDict' ) return serialized_info tensor_info = super(SequenceDict, self).get_serialized_info() return utils.map_nested(add_length_dim, tensor_info)
Returns SplitGenerators. def _split_generators(self, dl_manager): """Returns SplitGenerators.""" # Download the full MNIST Database filenames = { "train_data": _MNIST_TRAIN_DATA_FILENAME, "train_labels": _MNIST_TRAIN_LABELS_FILENAME, "test_data": _MNIST_TEST_DATA_FILENAME, "test_labels": _MNIST_TEST_LABELS_FILENAME, } mnist_files = dl_manager.download_and_extract( {k: urllib.parse.urljoin(self.URL, v) for k, v in filenames.items()}) # MNIST provides TRAIN and TEST splits, not a VALIDATION split, so we only # write the TRAIN and TEST splits to disk. return [ tfds.core.SplitGenerator( name=tfds.Split.TRAIN, num_shards=10, gen_kwargs=dict( num_examples=_TRAIN_EXAMPLES, data_path=mnist_files["train_data"], label_path=mnist_files["train_labels"], )), tfds.core.SplitGenerator( name=tfds.Split.TEST, num_shards=1, gen_kwargs=dict( num_examples=_TEST_EXAMPLES, data_path=mnist_files["test_data"], label_path=mnist_files["test_labels"], )), ]
Generate MNIST examples as dicts. Args: num_examples (int): The number of example. data_path (str): Path to the data files label_path (str): Path to the labels Yields: Generator yielding the next examples def _generate_examples(self, num_examples, data_path, label_path): """Generate MNIST examples as dicts. Args: num_examples (int): The number of example. data_path (str): Path to the data files label_path (str): Path to the labels Yields: Generator yielding the next examples """ images = _extract_mnist_images(data_path, num_examples) labels = _extract_mnist_labels(label_path, num_examples) data = list(zip(images, labels)) # Data is shuffled automatically to distribute classes uniformly. for image, label in data: yield { "image": image, "label": label, }
Returns SplitGenerators. def _split_generators(self, dl_manager): """Returns SplitGenerators.""" # Download images and annotations that come in separate archives. # Note, that the extension of archives is .tar.gz even though the actual # archives format is uncompressed tar. dl_paths = dl_manager.download_and_extract({ "images": tfds.download.Resource( url=os.path.join(_BASE_URL, "102flowers.tgz"), extract_method=tfds.download.ExtractMethod.TAR), "labels": os.path.join(_BASE_URL, "imagelabels.mat"), "setid": os.path.join(_BASE_URL, "setid.mat"), }) gen_kwargs = dict( images_dir_path=os.path.join(dl_paths["images"], "jpg"), labels_path=dl_paths["labels"], setid_path=dl_paths["setid"], ) return [ tfds.core.SplitGenerator( name=tfds.Split.TRAIN, num_shards=1, gen_kwargs=dict(split_name="trnid", **gen_kwargs)), tfds.core.SplitGenerator( name=tfds.Split.TEST, num_shards=1, gen_kwargs=dict(split_name="tstid", **gen_kwargs)), tfds.core.SplitGenerator( name=tfds.Split.VALIDATION, num_shards=1, gen_kwargs=dict(split_name="valid", **gen_kwargs)), ]
Yields examples. def _generate_examples(self, images_dir_path, labels_path, setid_path, split_name): """Yields examples.""" with tf.io.gfile.GFile(labels_path, "rb") as f: labels = tfds.core.lazy_imports.scipy.io.loadmat(f)["labels"][0] with tf.io.gfile.GFile(setid_path, "rb") as f: examples = tfds.core.lazy_imports.scipy.io.loadmat(f)[split_name][0] for image_id in examples: file_name = "image_%05d.jpg" % image_id yield { "image": os.path.join(images_dir_path, file_name), "label": labels[image_id - 1] - 1, "file_name": file_name, }
Calculate statistics for the specified split. def get_dataset_feature_statistics(builder, split): """Calculate statistics for the specified split.""" statistics = statistics_pb2.DatasetFeatureStatistics() # Make this to the best of our abilities. schema = schema_pb2.Schema() dataset = builder.as_dataset(split=split) # Just computing the number of examples for now. statistics.num_examples = 0 # Feature dictionaries. feature_to_num_examples = collections.defaultdict(int) feature_to_min = {} feature_to_max = {} np_dataset = dataset_utils.as_numpy(dataset) for example in utils.tqdm(np_dataset, unit=" examples", leave=False): statistics.num_examples += 1 assert isinstance(example, dict) feature_names = sorted(example.keys()) for feature_name in feature_names: # Update the number of examples this feature appears in. feature_to_num_examples[feature_name] += 1 feature_np = example[feature_name] # For compatibility in graph and eager mode, we can get PODs here and # everything may not be neatly wrapped up in numpy's ndarray. feature_dtype = type(feature_np) if isinstance(feature_np, np.ndarray): # If we have an empty array, then don't proceed further with computing # statistics on it. if feature_np.size == 0: continue feature_dtype = feature_np.dtype.type feature_min, feature_max = None, None is_numeric = (np.issubdtype(feature_dtype, np.number) or feature_dtype == np.bool_) if is_numeric: feature_min = np.min(feature_np) feature_max = np.max(feature_np) # TODO(afrozm): What if shapes don't match? Populate ValueCount? Add # logic for that. # Set or update the min, max. if is_numeric: if ((feature_name not in feature_to_min) or (feature_to_min[feature_name] > feature_min)): feature_to_min[feature_name] = feature_min if ((feature_name not in feature_to_max) or (feature_to_max[feature_name] < feature_max)): feature_to_max[feature_name] = feature_max # Start here, we've processed all examples. output_shapes_dict = dataset.output_shapes output_types_dict = dataset.output_types for feature_name in sorted(feature_to_num_examples.keys()): # Try to fill in the schema. feature = schema.feature.add() feature.name = feature_name # TODO(afrozm): Make this work with nested structures, currently the Schema # proto has no support for it. maybe_feature_shape = output_shapes_dict[feature_name] if not isinstance(maybe_feature_shape, tf.TensorShape): logging.error( "Statistics generation doesn't work for nested structures yet") continue for dim in maybe_feature_shape.as_list(): # We denote `None`s as -1 in the shape proto. feature.shape.dim.add().size = dim if dim else -1 feature_type = output_types_dict[feature_name] feature.type = _FEATURE_TYPE_MAP.get(feature_type, schema_pb2.BYTES) common_statistics = statistics_pb2.CommonStatistics() common_statistics.num_non_missing = feature_to_num_examples[feature_name] common_statistics.num_missing = ( statistics.num_examples - common_statistics.num_non_missing) feature_name_statistics = statistics.features.add() feature_name_statistics.name = feature_name # TODO(afrozm): This can be skipped, since type information was added to # the Schema. feature_name_statistics.type = _SCHEMA_TYPE_MAP.get( feature.type, statistics_pb2.FeatureNameStatistics.BYTES) if feature.type == schema_pb2.INT or feature.type == schema_pb2.FLOAT: numeric_statistics = statistics_pb2.NumericStatistics() numeric_statistics.min = feature_to_min[feature_name] numeric_statistics.max = feature_to_max[feature_name] numeric_statistics.common_stats.CopyFrom(common_statistics) feature_name_statistics.num_stats.CopyFrom(numeric_statistics) else: # Let's shove it into BytesStatistics for now. bytes_statistics = statistics_pb2.BytesStatistics() bytes_statistics.common_stats.CopyFrom(common_statistics) feature_name_statistics.bytes_stats.CopyFrom(bytes_statistics) return statistics, schema
Read JSON-formatted proto into DatasetInfo proto. def read_from_json(json_filename): """Read JSON-formatted proto into DatasetInfo proto.""" with tf.io.gfile.GFile(json_filename) as f: dataset_info_json_str = f.read() # Parse it back into a proto. parsed_proto = json_format.Parse(dataset_info_json_str, dataset_info_pb2.DatasetInfo()) return parsed_proto
Full canonical name: (<dataset_name>/<config_name>/<version>). def full_name(self): """Full canonical name: (<dataset_name>/<config_name>/<version>).""" names = [self._builder.name] if self._builder.builder_config: names.append(self._builder.builder_config.name) names.append(str(self.version)) return posixpath.join(*names)
Overwrite the splits if they are different from the current ones. * If splits aren't already defined or different (ex: different number of shards), then the new split dict is used. This will trigger stats computation during download_and_prepare. * If splits are already defined in DatasetInfo and similar (same names and shards): keep the restored split which contains the statistics (restored from GCS or file) Args: split_dict: `tfds.core.SplitDict`, the new split def update_splits_if_different(self, split_dict): """Overwrite the splits if they are different from the current ones. * If splits aren't already defined or different (ex: different number of shards), then the new split dict is used. This will trigger stats computation during download_and_prepare. * If splits are already defined in DatasetInfo and similar (same names and shards): keep the restored split which contains the statistics (restored from GCS or file) Args: split_dict: `tfds.core.SplitDict`, the new split """ assert isinstance(split_dict, splits_lib.SplitDict) # If splits are already defined and identical, then we do not update if self._splits and splits_lib.check_splits_equals( self._splits, split_dict): return self._set_splits(split_dict)
Split setter (private method). def _set_splits(self, split_dict): """Split setter (private method).""" # Update the dictionary representation. # Use from/to proto for a clean copy self._splits = split_dict.copy() # Update the proto del self.as_proto.splits[:] # Clear previous for split_info in split_dict.to_proto(): self.as_proto.splits.add().CopyFrom(split_info)
Update from the DatasetBuilder. def _compute_dynamic_properties(self, builder): """Update from the DatasetBuilder.""" # Fill other things by going over the dataset. splits = self.splits for split_info in utils.tqdm( splits.values(), desc="Computing statistics...", unit=" split"): try: split_name = split_info.name # Fill DatasetFeatureStatistics. dataset_feature_statistics, schema = get_dataset_feature_statistics( builder, split_name) # Add the statistics to this split. split_info.statistics.CopyFrom(dataset_feature_statistics) # Set the schema at the top-level since this is independent of the # split. self.as_proto.schema.CopyFrom(schema) except tf.errors.InvalidArgumentError: # This means there is no such split, even though it was specified in the # info, the least we can do is to log this. logging.error(("%s's info() property specifies split %s, but it " "doesn't seem to have been generated. Please ensure " "that the data was downloaded for this split and re-run " "download_and_prepare."), self.name, split_name) raise # Set splits to trigger proto update in setter self._set_splits(splits)
Write `DatasetInfo` as JSON to `dataset_info_dir`. def write_to_directory(self, dataset_info_dir): """Write `DatasetInfo` as JSON to `dataset_info_dir`.""" # Save the metadata from the features (vocabulary, labels,...) if self.features: self.features.save_metadata(dataset_info_dir) if self.redistribution_info.license: with tf.io.gfile.GFile(self._license_filename(dataset_info_dir), "w") as f: f.write(self.redistribution_info.license) with tf.io.gfile.GFile(self._dataset_info_filename(dataset_info_dir), "w") as f: f.write(self.as_json)
Update DatasetInfo from the JSON file in `dataset_info_dir`. This function updates all the dynamically generated fields (num_examples, hash, time of creation,...) of the DatasetInfo. This will overwrite all previous metadata. Args: dataset_info_dir: `str` The directory containing the metadata file. This should be the root directory of a specific dataset version. def read_from_directory(self, dataset_info_dir): """Update DatasetInfo from the JSON file in `dataset_info_dir`. This function updates all the dynamically generated fields (num_examples, hash, time of creation,...) of the DatasetInfo. This will overwrite all previous metadata. Args: dataset_info_dir: `str` The directory containing the metadata file. This should be the root directory of a specific dataset version. """ if not dataset_info_dir: raise ValueError( "Calling read_from_directory with undefined dataset_info_dir.") json_filename = self._dataset_info_filename(dataset_info_dir) # Load the metadata from disk parsed_proto = read_from_json(json_filename) # Update splits self._set_splits(splits_lib.SplitDict.from_proto(parsed_proto.splits)) # Restore the feature metadata (vocabulary, labels names,...) if self.features: self.features.load_metadata(dataset_info_dir) # Update fields which are not defined in the code. This means that # the code will overwrite fields which are present in # dataset_info.json. for field_name, field in self.as_proto.DESCRIPTOR.fields_by_name.items(): field_value = getattr(self._info_proto, field_name) field_value_restored = getattr(parsed_proto, field_name) try: is_defined = self._info_proto.HasField(field_name) except ValueError: is_defined = bool(field_value) try: is_defined_in_restored = parsed_proto.HasField(field_name) except ValueError: is_defined_in_restored = bool(field_value_restored) # If field is defined in code, we ignore the value if is_defined: if field_value != field_value_restored: logging.info( "Field info.%s from disk and from code do not match. Keeping " "the one from code.", field_name) continue # If the field is also not defined in JSON file, we do nothing if not is_defined_in_restored: continue # Otherwise, we restore the dataset_info.json value if field.type == field.TYPE_MESSAGE: field_value.MergeFrom(field_value_restored) else: setattr(self._info_proto, field_name, field_value_restored) if self._builder._version != self.version: # pylint: disable=protected-access raise AssertionError( "The constructed DatasetInfo instance and the restored proto version " "do not match. Builder version: {}. Proto version: {}".format( self._builder._version, self.version)) # pylint: disable=protected-access # Mark as fully initialized. self._fully_initialized = True
Initialize DatasetInfo from GCS bucket info files. def initialize_from_bucket(self): """Initialize DatasetInfo from GCS bucket info files.""" # In order to support Colab, we use the HTTP GCS API to access the metadata # files. They are copied locally and then loaded. tmp_dir = tempfile.mkdtemp("tfds") data_files = gcs_utils.gcs_dataset_info_files(self.full_name) if not data_files: return logging.info("Loading info from GCS for %s", self.full_name) for fname in data_files: out_fname = os.path.join(tmp_dir, os.path.basename(fname)) gcs_utils.download_gcs_file(fname, out_fname) self.read_from_directory(tmp_dir)
Returns SplitGenerators. def _split_generators(self, dl_manager): """Returns SplitGenerators.""" url = _DL_URLS[self.builder_config.name] data_dirs = dl_manager.download_and_extract(url) path_to_dataset = os.path.join(data_dirs, tf.io.gfile.listdir(data_dirs)[0]) train_a_path = os.path.join(path_to_dataset, "trainA") train_b_path = os.path.join(path_to_dataset, "trainB") test_a_path = os.path.join(path_to_dataset, "testA") test_b_path = os.path.join(path_to_dataset, "testB") return [ tfds.core.SplitGenerator( name="trainA", num_shards=10, gen_kwargs={ "path": train_a_path, "label": "A", }), tfds.core.SplitGenerator( name="trainB", num_shards=10, gen_kwargs={ "path": train_b_path, "label": "B", }), tfds.core.SplitGenerator( name="testA", num_shards=1, gen_kwargs={ "path": test_a_path, "label": "A", }), tfds.core.SplitGenerator( name="testB", num_shards=1, gen_kwargs={ "path": test_b_path, "label": "B", }), ]
Map the function into each element and resolve the promise. def _map_promise(map_fn, all_inputs): """Map the function into each element and resolve the promise.""" all_promises = utils.map_nested(map_fn, all_inputs) # Apply the function res = utils.map_nested(_wait_on_promise, all_promises) return res
Store dled file to definitive place, write INFO file, return path. def _handle_download_result(self, resource, tmp_dir_path, sha256, dl_size): """Store dled file to definitive place, write INFO file, return path.""" fnames = tf.io.gfile.listdir(tmp_dir_path) if len(fnames) > 1: raise AssertionError('More than one file in %s.' % tmp_dir_path) original_fname = fnames[0] tmp_path = os.path.join(tmp_dir_path, original_fname) self._recorded_sizes_checksums[resource.url] = (dl_size, sha256) if self._register_checksums: self._record_sizes_checksums() elif (dl_size, sha256) != self._sizes_checksums.get(resource.url, None): raise NonMatchingChecksumError(resource.url, tmp_path) download_path = self._get_final_dl_path(resource.url, sha256) resource_lib.write_info_file(resource, download_path, self._dataset_name, original_fname) # Unconditionally overwrite because either file doesn't exist or # FORCE_DOWNLOAD=true tf.io.gfile.rename(tmp_path, download_path, overwrite=True) tf.io.gfile.rmtree(tmp_dir_path) return download_path
Download resource, returns Promise->path to downloaded file. def _download(self, resource): """Download resource, returns Promise->path to downloaded file.""" if isinstance(resource, six.string_types): resource = resource_lib.Resource(url=resource) url = resource.url if url in self._sizes_checksums: expected_sha256 = self._sizes_checksums[url][1] download_path = self._get_final_dl_path(url, expected_sha256) if not self._force_download and resource.exists_locally(download_path): logging.info('URL %s already downloaded: reusing %s.', url, download_path) self._recorded_sizes_checksums[url] = self._sizes_checksums[url] return promise.Promise.resolve(download_path) # There is a slight difference between downloader and extractor here: # the extractor manages its own temp directory, while the DownloadManager # manages the temp directory of downloader. download_dir_path = os.path.join( self._download_dir, '%s.tmp.%s' % (resource_lib.get_dl_dirname(url), uuid.uuid4().hex)) tf.io.gfile.makedirs(download_dir_path) logging.info('Downloading %s into %s...', url, download_dir_path) def callback(val): checksum, dl_size = val return self._handle_download_result( resource, download_dir_path, checksum, dl_size) return self._downloader.download(url, download_dir_path).then(callback)
Extract a single archive, returns Promise->path to extraction result. def _extract(self, resource): """Extract a single archive, returns Promise->path to extraction result.""" if isinstance(resource, six.string_types): resource = resource_lib.Resource(path=resource) path = resource.path extract_method = resource.extract_method if extract_method == resource_lib.ExtractMethod.NO_EXTRACT: logging.info('Skipping extraction for %s (method=NO_EXTRACT).', path) return promise.Promise.resolve(path) method_name = resource_lib.ExtractMethod(extract_method).name extract_path = os.path.join(self._extract_dir, '%s.%s' % (method_name, os.path.basename(path))) if not self._force_extraction and tf.io.gfile.exists(extract_path): logging.info('Reusing extraction of %s at %s.', path, extract_path) return promise.Promise.resolve(extract_path) return self._extractor.extract(path, extract_method, extract_path)
Download-extract `Resource` or url, returns Promise->path. def _download_extract(self, resource): """Download-extract `Resource` or url, returns Promise->path.""" if isinstance(resource, six.string_types): resource = resource_lib.Resource(url=resource) def callback(path): resource.path = path return self._extract(resource) return self._download(resource).then(callback)
Download data for a given Kaggle competition. def download_kaggle_data(self, competition_name): """Download data for a given Kaggle competition.""" with self._downloader.tqdm(): kaggle_downloader = self._downloader.kaggle_downloader(competition_name) urls = kaggle_downloader.competition_urls files = kaggle_downloader.competition_files return _map_promise(self._download, dict((f, u) for (f, u) in zip(files, urls)))
Download given url(s). Args: url_or_urls: url or `list`/`dict` of urls to download and extract. Each url can be a `str` or `tfds.download.Resource`. Returns: downloaded_path(s): `str`, The downloaded paths matching the given input url_or_urls. def download(self, url_or_urls): """Download given url(s). Args: url_or_urls: url or `list`/`dict` of urls to download and extract. Each url can be a `str` or `tfds.download.Resource`. Returns: downloaded_path(s): `str`, The downloaded paths matching the given input url_or_urls. """ # Add progress bar to follow the download state with self._downloader.tqdm(): return _map_promise(self._download, url_or_urls)
Returns iterator over files within archive. **Important Note**: caller should read files as they are yielded. Reading out of order is slow. Args: resource: path to archive or `tfds.download.Resource`. Returns: Generator yielding tuple (path_within_archive, file_obj). def iter_archive(self, resource): """Returns iterator over files within archive. **Important Note**: caller should read files as they are yielded. Reading out of order is slow. Args: resource: path to archive or `tfds.download.Resource`. Returns: Generator yielding tuple (path_within_archive, file_obj). """ if isinstance(resource, six.string_types): resource = resource_lib.Resource(path=resource) return extractor.iter_archive(resource.path, resource.extract_method)
Extract given path(s). Args: path_or_paths: path or `list`/`dict` of path of file to extract. Each path can be a `str` or `tfds.download.Resource`. If not explicitly specified in `Resource`, the extraction method is deduced from downloaded file name. Returns: extracted_path(s): `str`, The extracted paths matching the given input path_or_paths. def extract(self, path_or_paths): """Extract given path(s). Args: path_or_paths: path or `list`/`dict` of path of file to extract. Each path can be a `str` or `tfds.download.Resource`. If not explicitly specified in `Resource`, the extraction method is deduced from downloaded file name. Returns: extracted_path(s): `str`, The extracted paths matching the given input path_or_paths. """ # Add progress bar to follow the download state with self._extractor.tqdm(): return _map_promise(self._extract, path_or_paths)
Download and extract given url_or_urls. Is roughly equivalent to: ``` extracted_paths = dl_manager.extract(dl_manager.download(url_or_urls)) ``` Args: url_or_urls: url or `list`/`dict` of urls to download and extract. Each url can be a `str` or `tfds.download.Resource`. If not explicitly specified in `Resource`, the extraction method will automatically be deduced from downloaded file name. Returns: extracted_path(s): `str`, extracted paths of given URL(s). def download_and_extract(self, url_or_urls): """Download and extract given url_or_urls. Is roughly equivalent to: ``` extracted_paths = dl_manager.extract(dl_manager.download(url_or_urls)) ``` Args: url_or_urls: url or `list`/`dict` of urls to download and extract. Each url can be a `str` or `tfds.download.Resource`. If not explicitly specified in `Resource`, the extraction method will automatically be deduced from downloaded file name. Returns: extracted_path(s): `str`, extracted paths of given URL(s). """ # Add progress bar to follow the download state with self._downloader.tqdm(): with self._extractor.tqdm(): return _map_promise(self._download_extract, url_or_urls)
Returns the directory containing the manually extracted data. def manual_dir(self): """Returns the directory containing the manually extracted data.""" if not tf.io.gfile.exists(self._manual_dir): raise AssertionError( 'Manual directory {} does not exist. Create it and download/extract ' 'dataset artifacts in there.'.format(self._manual_dir)) return self._manual_dir
Construct a list of BuilderConfigs. Construct a list of 75 Cifar10CorruptedConfig objects, corresponding to the 15 corruption types and 5 severities. Returns: A list of 75 Cifar10CorruptedConfig objects. def _make_builder_configs(): """Construct a list of BuilderConfigs. Construct a list of 75 Cifar10CorruptedConfig objects, corresponding to the 15 corruption types and 5 severities. Returns: A list of 75 Cifar10CorruptedConfig objects. """ config_list = [] for corruption in _CORRUPTIONS: for severity in range(1, 6): config_list.append( Cifar10CorruptedConfig( name=corruption + '_' + str(severity), version='0.0.1', description='Corruption method: ' + corruption + ', severity level: ' + str(severity), corruption_type=corruption, severity=severity, )) return config_list
Return the test split of Cifar10. Args: dl_manager: download manager object. Returns: test split. def _split_generators(self, dl_manager): """Return the test split of Cifar10. Args: dl_manager: download manager object. Returns: test split. """ path = dl_manager.download_and_extract(_DOWNLOAD_URL) return [ tfds.core.SplitGenerator( name=tfds.Split.TEST, num_shards=1, gen_kwargs={'data_dir': os.path.join(path, _DIRNAME)}) ]
Generate corrupted Cifar10 test data. Apply corruptions to the raw images according to self.corruption_type. Args: data_dir: root directory of downloaded dataset Yields: dictionary with image file and label. def _generate_examples(self, data_dir): """Generate corrupted Cifar10 test data. Apply corruptions to the raw images according to self.corruption_type. Args: data_dir: root directory of downloaded dataset Yields: dictionary with image file and label. """ corruption = self.builder_config.corruption severity = self.builder_config.severity images_file = os.path.join(data_dir, _CORRUPTIONS_TO_FILENAMES[corruption]) labels_file = os.path.join(data_dir, _LABELS_FILENAME) with tf.io.gfile.GFile(labels_file, mode='rb') as f: labels = np.load(f) num_images = labels.shape[0] // 5 # Labels are stacked 5 times so we can just read the first iteration labels = labels[:num_images] with tf.io.gfile.GFile(images_file, mode='rb') as f: images = np.load(f) # Slice images corresponding to correct severity level images = images[(severity - 1) * num_images:severity * num_images] for image, label in zip(images, labels): yield { 'image': image, 'label': label, }
Doc string for a single builder, with or without configs. def document_single_builder(builder): """Doc string for a single builder, with or without configs.""" mod_name = builder.__class__.__module__ cls_name = builder.__class__.__name__ mod_file = sys.modules[mod_name].__file__ if mod_file.endswith("pyc"): mod_file = mod_file[:-1] description_prefix = "" if builder.builder_configs: # Dataset with configs; document each one config_docs = [] for config in builder.BUILDER_CONFIGS: builder = tfds.builder(builder.name, config=config) info = builder.info # TODO(rsepassi): document the actual config object config_doc = SINGLE_CONFIG_ENTRY.format( builder_name=builder.name, config_name=config.name, description=config.description, version=config.version, feature_information=make_feature_information(info), size=tfds.units.size_str(info.size_in_bytes), ) config_docs.append(config_doc) out_str = DATASET_WITH_CONFIGS_ENTRY.format( snakecase_name=builder.name, module_and_class="%s.%s" % (tfds_mod_name(mod_name), cls_name), cls_url=cls_url(mod_name), config_names="\n".join([ CONFIG_BULLET.format(name=config.name, description=config.description, version=config.version, size=tfds.units.size_str(tfds.builder( builder.name, config=config) .info.size_in_bytes)) for config in builder.BUILDER_CONFIGS]), config_cls="%s.%s" % (tfds_mod_name(mod_name), type(builder.builder_config).__name__), configs="\n".join(config_docs), urls=format_urls(info.urls), url=url_from_info(info), supervised_keys=str(info.supervised_keys), citation=make_citation(info.citation), statistics_information=make_statistics_information(info), description=builder.info.description, description_prefix=description_prefix, ) else: info = builder.info out_str = DATASET_ENTRY.format( snakecase_name=builder.name, module_and_class="%s.%s" % (tfds_mod_name(mod_name), cls_name), cls_url=cls_url(mod_name), description=info.description, description_prefix=description_prefix, version=info.version, feature_information=make_feature_information(info), statistics_information=make_statistics_information(info), urls=format_urls(info.urls), url=url_from_info(info), supervised_keys=str(info.supervised_keys), citation=make_citation(info.citation), size=tfds.units.size_str(info.size_in_bytes), ) out_str = schema_org(builder) + "\n" + out_str return out_str
Get all builders organized by module in nested dicts. def make_module_to_builder_dict(datasets=None): """Get all builders organized by module in nested dicts.""" # pylint: disable=g-long-lambda # dict to hold tfds->image->mnist->[builders] module_to_builder = collections.defaultdict( lambda: collections.defaultdict( lambda: collections.defaultdict(list))) # pylint: enable=g-long-lambda if datasets: builders = [tfds.builder(name) for name in datasets] else: builders = [ tfds.builder(name) for name in tfds.list_builders() if name not in BUILDER_BLACKLIST ] + [tfds.builder("image_label_folder", dataset_name="image_label_folder")] for builder in builders: mod_name = builder.__class__.__module__ modules = mod_name.split(".") if "testing" in modules: continue current_mod_ctr = module_to_builder for mod in modules: current_mod_ctr = current_mod_ctr[mod] current_mod_ctr.append(builder) module_to_builder = module_to_builder["tensorflow_datasets"] return module_to_builder
Pretty-print tfds.features.FeaturesDict. def _pprint_features_dict(features_dict, indent=0, add_prefix=True): """Pretty-print tfds.features.FeaturesDict.""" first_last_indent_str = " " * indent indent_str = " " * (indent + 4) first_line = "%s%s({" % ( first_last_indent_str if add_prefix else "", type(features_dict).__name__, ) lines = [first_line] for k in sorted(list(features_dict.keys())): v = features_dict[k] if isinstance(v, tfds.features.FeaturesDict): v_str = _pprint_features_dict(v, indent + 4, False) else: v_str = str(v) lines.append("%s'%s': %s," % (indent_str, k, v_str)) lines.append("%s})" % first_last_indent_str) return "\n".join(lines)
Make statistics information table. def make_statistics_information(info): """Make statistics information table.""" if not info.splits.total_num_examples: # That means that we have yet to calculate the statistics for this. return "None computed" stats = [(info.splits.total_num_examples, "ALL")] for split_name, split_info in info.splits.items(): stats.append((split_info.num_examples, split_name.upper())) # Sort reverse on number of examples. stats.sort(reverse=True) stats = "\n".join([ "{0:10} | {1:>10,}".format(name, num_exs) for (num_exs, name) in stats ]) return STATISTICS_TABLE.format(split_statistics=stats)
Create dataset documentation string for given datasets. Args: datasets: list of datasets for which to create documentation. If None, then all available datasets will be used. Returns: string describing the datasets (in the MarkDown format). def dataset_docs_str(datasets=None): """Create dataset documentation string for given datasets. Args: datasets: list of datasets for which to create documentation. If None, then all available datasets will be used. Returns: string describing the datasets (in the MarkDown format). """ module_to_builder = make_module_to_builder_dict(datasets) sections = sorted(list(module_to_builder.keys())) section_tocs = [] section_docs = [] for section in sections: builders = tf.nest.flatten(module_to_builder[section]) builders = sorted(builders, key=lambda b: b.name) builder_docs = [document_single_builder(builder) for builder in builders] section_doc = SECTION_DATASETS.format( section_name=section, datasets="\n".join(builder_docs)) section_toc = create_section_toc(section, builders) section_docs.append(section_doc) section_tocs.append(section_toc) full_doc = DOC.format(toc="\n".join(section_tocs), datasets="\n".join(section_docs)) return full_doc
Builds schema.org microdata for DatasetSearch from DatasetBuilder. Markup spec: https://developers.google.com/search/docs/data-types/dataset#dataset Testing tool: https://search.google.com/structured-data/testing-tool For Google Dataset Search: https://toolbox.google.com/datasetsearch Microdata format was chosen over JSON-LD due to the fact that Markdown rendering engines remove all <script> tags. Args: builder: `tfds.core.DatasetBuilder` Returns: HTML string with microdata def schema_org(builder): # pylint: disable=line-too-long """Builds schema.org microdata for DatasetSearch from DatasetBuilder. Markup spec: https://developers.google.com/search/docs/data-types/dataset#dataset Testing tool: https://search.google.com/structured-data/testing-tool For Google Dataset Search: https://toolbox.google.com/datasetsearch Microdata format was chosen over JSON-LD due to the fact that Markdown rendering engines remove all <script> tags. Args: builder: `tfds.core.DatasetBuilder` Returns: HTML string with microdata """ # pylint: enable=line-too-long properties = [ (lambda x: x.name, SCHEMA_ORG_NAME), (lambda x: x.description, SCHEMA_ORG_DESC), (lambda x: x.name, SCHEMA_ORG_URL), (lambda x: (x.urls and x.urls[0]) or "", SCHEMA_ORG_SAMEAS) ] info = builder.info out_str = SCHEMA_ORG_PRE for extractor, template in properties: val = extractor(info) if val: # We are using cgi module instead of html due to Python 2 compatibility out_str += template.format(val=cgi.escape(val, quote=True).strip()) out_str += SCHEMA_ORG_POST return out_str
Generating a Gaussian blurring kernel with disk shape. Generating a Gaussian blurring kernel with disk shape using cv2 API. Args: radius: integer, radius of blurring kernel. alias_blur: float, standard deviation of Gaussian blurring. dtype: data type of kernel Returns: cv2 object of the Gaussian blurring kernel. def disk(radius, alias_blur=0.1, dtype=np.float32): """Generating a Gaussian blurring kernel with disk shape. Generating a Gaussian blurring kernel with disk shape using cv2 API. Args: radius: integer, radius of blurring kernel. alias_blur: float, standard deviation of Gaussian blurring. dtype: data type of kernel Returns: cv2 object of the Gaussian blurring kernel. """ if radius <= 8: length = np.arange(-8, 8 + 1) ksize = (3, 3) else: length = np.arange(-radius, radius + 1) ksize = (5, 5) x_axis, y_axis = np.meshgrid(length, length) aliased_disk = np.array((x_axis**2 + y_axis**2) <= radius**2, dtype=dtype) aliased_disk /= np.sum(aliased_disk) # supersample disk to antialias return tfds.core.lazy_imports.cv2.GaussianBlur( aliased_disk, ksize=ksize, sigmaX=alias_blur)
Zoom image with clipping. Zoom the central part of the image and clip extra pixels. Args: img: numpy array, uncorrupted image. zoom_factor: numpy array, a sequence of float numbers for zoom factor. Returns: numpy array, zoomed image after clipping. def clipped_zoom(img, zoom_factor): """Zoom image with clipping. Zoom the central part of the image and clip extra pixels. Args: img: numpy array, uncorrupted image. zoom_factor: numpy array, a sequence of float numbers for zoom factor. Returns: numpy array, zoomed image after clipping. """ h = img.shape[0] ch = int(np.ceil(h / float(zoom_factor))) top_h = (h - ch) // 2 w = img.shape[1] cw = int(np.ceil(w / float(zoom_factor))) top_w = (w - cw) // 2 img = tfds.core.lazy_imports.scipy.ndimage.zoom( img[top_h:top_h + ch, top_w:top_w + cw], (zoom_factor, zoom_factor, 1), order=1) # trim off any extra pixels trim_top_h = (img.shape[0] - h) // 2 trim_top_w = (img.shape[1] - w) // 2 return img[trim_top_h:trim_top_h + h, trim_top_w:trim_top_w + w]
Generate a heightmap using diamond-square algorithm. Modification of the algorithm in https://github.com/FLHerne/mapgen/blob/master/diamondsquare.py Args: mapsize: side length of the heightmap, must be a power of two. wibbledecay: integer, decay factor. Returns: numpy 2d array, side length 'mapsize', of floats in [0,255]. def plasma_fractal(mapsize=512, wibbledecay=3): """Generate a heightmap using diamond-square algorithm. Modification of the algorithm in https://github.com/FLHerne/mapgen/blob/master/diamondsquare.py Args: mapsize: side length of the heightmap, must be a power of two. wibbledecay: integer, decay factor. Returns: numpy 2d array, side length 'mapsize', of floats in [0,255]. """ if mapsize & (mapsize - 1) != 0: raise ValueError('mapsize must be a power of two.') maparray = np.empty((mapsize, mapsize), dtype=np.float_) maparray[0, 0] = 0 stepsize = mapsize wibble = 100 def wibbledmean(array): return array / 4 + wibble * np.random.uniform(-wibble, wibble, array.shape) def fillsquares(): """For each square, calculate middle value as mean of points + wibble.""" cornerref = maparray[0:mapsize:stepsize, 0:mapsize:stepsize] squareaccum = cornerref + np.roll(cornerref, shift=-1, axis=0) squareaccum += np.roll(squareaccum, shift=-1, axis=1) maparray[stepsize // 2:mapsize:stepsize, stepsize // 2:mapsize:stepsize] = wibbledmean(squareaccum) def filldiamonds(): """For each diamond, calculate middle value as meanof points + wibble.""" mapsize = maparray.shape[0] drgrid = maparray[stepsize // 2:mapsize:stepsize, stepsize // 2:mapsize:stepsize] ulgrid = maparray[0:mapsize:stepsize, 0:mapsize:stepsize] ldrsum = drgrid + np.roll(drgrid, 1, axis=0) lulsum = ulgrid + np.roll(ulgrid, -1, axis=1) ltsum = ldrsum + lulsum maparray[0:mapsize:stepsize, stepsize // 2:mapsize:stepsize] = wibbledmean(ltsum) tdrsum = drgrid + np.roll(drgrid, 1, axis=1) tulsum = ulgrid + np.roll(ulgrid, -1, axis=0) ttsum = tdrsum + tulsum maparray[stepsize // 2:mapsize:stepsize, 0:mapsize:stepsize] = wibbledmean(ttsum) while stepsize >= 2: fillsquares() filldiamonds() stepsize //= 2 wibble /= wibbledecay maparray -= maparray.min() return maparray / maparray.max()
Gaussian noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added Gaussian noise. def gaussian_noise(x, severity=1): """Gaussian noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added Gaussian noise. """ c = [.08, .12, 0.18, 0.26, 0.38][severity - 1] x = np.array(x) / 255. x_clip = np.clip(x + np.random.normal(size=x.shape, scale=c), 0, 1) * 255 return around_and_astype(x_clip)
Shot noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added shot noise. def shot_noise(x, severity=1): """Shot noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added shot noise. """ c = [60, 25, 12, 5, 3][severity - 1] x = np.array(x) / 255. x_clip = np.clip(np.random.poisson(x * c) / float(c), 0, 1) * 255 return around_and_astype(x_clip)
Impulse noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added impulse noise. def impulse_noise(x, severity=1): """Impulse noise corruption to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added impulse noise. """ c = [.03, .06, .09, 0.17, 0.27][severity - 1] x = tfds.core.lazy_imports.skimage.util.random_noise( np.array(x) / 255., mode='s&p', amount=c) x_clip = np.clip(x, 0, 1) * 255 return around_and_astype(x_clip)
Defocus blurring to images. Apply defocus blurring to images using Gaussian kernel. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied defocus blur. def defocus_blur(x, severity=1): """Defocus blurring to images. Apply defocus blurring to images using Gaussian kernel. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied defocus blur. """ c = [(3, 0.1), (4, 0.5), (6, 0.5), (8, 0.5), (10, 0.5)][severity - 1] x = np.array(x) / 255. kernel = disk(radius=c[0], alias_blur=c[1]) channels = [] for d in range(3): channels.append(tfds.core.lazy_imports.cv2.filter2D(x[:, :, d], -1, kernel)) channels = np.array(channels).transpose((1, 2, 0)) # 3x224x224 -> 224x224x3 x_clip = np.clip(channels, 0, 1) * 255 return around_and_astype(x_clip)
Frosted glass blurring to images. Apply frosted glass blurring to images by shuffling pixels locally. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied frosted glass blur. def frosted_glass_blur(x, severity=1): """Frosted glass blurring to images. Apply frosted glass blurring to images by shuffling pixels locally. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied frosted glass blur. """ # sigma, max_delta, iterations c = [(0.7, 1, 2), (0.9, 2, 1), (1, 2, 3), (1.1, 3, 2), (1.5, 4, 2)][severity - 1] x = np.uint8( tfds.core.lazy_imports.skimage.filters.gaussian( np.array(x) / 255., sigma=c[0], multichannel=True) * 255) # locally shuffle pixels for _ in range(c[2]): for h in range(x.shape[0] - c[1], c[1], -1): for w in range(x.shape[1] - c[1], c[1], -1): dx, dy = np.random.randint(-c[1], c[1], size=(2,)) h_prime, w_prime = h + dy, w + dx # swap x[h, w], x[h_prime, w_prime] = x[h_prime, w_prime], x[h, w] x_clip = np.clip( tfds.core.lazy_imports.skimage.filters.gaussian( x / 255., sigma=c[0], multichannel=True), 0, 1) x_clip *= 255 return around_and_astype(x_clip)
Zoom blurring to images. Applying zoom blurring to images by zooming the central part of the images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied zoom blur. def zoom_blur(x, severity=1): """Zoom blurring to images. Applying zoom blurring to images by zooming the central part of the images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied zoom blur. """ c = [ np.arange(1, 1.11, 0.01), np.arange(1, 1.16, 0.01), np.arange(1, 1.21, 0.02), np.arange(1, 1.26, 0.02), np.arange(1, 1.31, 0.03) ][severity - 1] x = (np.array(x) / 255.).astype(np.float32) out = np.zeros_like(x) for zoom_factor in c: out += clipped_zoom(x, zoom_factor) x = (x + out) / (len(c) + 1) x_clip = np.clip(x, 0, 1) * 255 return around_and_astype(x_clip)
Fog corruption to images. Adding fog to images. Fog is generated by diamond-square algorithm. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added fog. def fog(x, severity=1): """Fog corruption to images. Adding fog to images. Fog is generated by diamond-square algorithm. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Added fog. """ c = [(1.5, 2), (2., 2), (2.5, 1.7), (2.5, 1.5), (3., 1.4)][severity - 1] x = np.array(x) / 255. max_val = x.max() mapsize = 512 shape = x.shape max_length = max(shape[0], shape[1]) if max_length > mapsize: mapsize = 2**int(np.ceil(np.log2(float(max_length)))) tmp = plasma_fractal(mapsize=mapsize, wibbledecay=c[1]) tmp = tmp[:x.shape[0], :x.shape[1]] tmp = tmp[..., np.newaxis] x += c[0] * tmp x_clip = np.clip(x * max_val / (max_val + c[0]), 0, 1) * 255 return around_and_astype(x_clip)
Change brightness of images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Changed brightness. def brightness(x, severity=1): """Change brightness of images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Changed brightness. """ c = [.1, .2, .3, .4, .5][severity - 1] x = np.array(x) / 255. x = tfds.core.lazy_imports.skimage.color.rgb2hsv(x) x[:, :, 2] = np.clip(x[:, :, 2] + c, 0, 1) x = tfds.core.lazy_imports.skimage.color.hsv2rgb(x) x_clip = np.clip(x, 0, 1) * 255 return around_and_astype(x_clip)
Change contrast of images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Changed contrast. def contrast(x, severity=1): """Change contrast of images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Changed contrast. """ c = [0.4, .3, .2, .1, .05][severity - 1] x = np.array(x) / 255. means = np.mean(x, axis=(0, 1), keepdims=True) x_clip = np.clip((x - means) * c + means, 0, 1) * 255 return around_and_astype(x_clip)
Conduct elastic transform to images. Elastic transform is performed on small patches of the images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied elastic transform. def elastic(x, severity=1): """Conduct elastic transform to images. Elastic transform is performed on small patches of the images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied elastic transform. """ c = [(244 * 2, 244 * 0.7, 244 * 0.1), (244 * 2, 244 * 0.08, 244 * 0.2), (244 * 0.05, 244 * 0.01, 244 * 0.02), (244 * 0.07, 244 * 0.01, 244 * 0.02), (244 * 0.12, 244 * 0.01, 244 * 0.02)][severity - 1] image = np.array(x, dtype=np.float32) / 255. shape = image.shape shape_size = shape[:2] # random affine center_square = np.float32(shape_size) // 2 square_size = min(shape_size) // 3 pts1 = np.float32([ center_square + square_size, [center_square[0] + square_size, center_square[1] - square_size], center_square - square_size ]) pts2 = pts1 + np.random.uniform( -c[2], c[2], size=pts1.shape).astype(np.float32) affine_trans = tfds.core.lazy_imports.cv2.getAffineTransform(pts1, pts2) image = tfds.core.lazy_imports.cv2.warpAffine( image, affine_trans, shape_size[::-1], borderMode=tfds.core.lazy_imports.cv2.BORDER_REFLECT_101) dx = (tfds.core.lazy_imports.skimage.filters.gaussian( np.random.uniform(-1, 1, size=shape[:2]), c[1], mode='reflect', truncate=3) * c[0]).astype(np.float32) dy = (tfds.core.lazy_imports.skimage.filters.gaussian( np.random.uniform(-1, 1, size=shape[:2]), c[1], mode='reflect', truncate=3) * c[0]).astype(np.float32) dx, dy = dx[..., np.newaxis], dy[..., np.newaxis] x, y, z = np.meshgrid( np.arange(shape[1]), np.arange(shape[0]), np.arange(shape[2])) indices = np.reshape(y + dy, (-1, 1)), np.reshape(x + dx, (-1, 1)), np.reshape( z, (-1, 1)) x_clip = np.clip( tfds.core.lazy_imports.scipy.ndimage.interpolation.map_coordinates( image, indices, order=1, mode='reflect').reshape(shape), 0, 1) * 255 return around_and_astype(x_clip)
Pixelate images. Conduct pixelating corruptions to images by first shrinking the images and then resizing to original size. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied pixelating corruption. def pixelate(x, severity=1): """Pixelate images. Conduct pixelating corruptions to images by first shrinking the images and then resizing to original size. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied pixelating corruption. """ c = [0.6, 0.5, 0.4, 0.3, 0.25][severity - 1] shape = x.shape x = tfds.core.lazy_imports.PIL_Image.fromarray(x.astype(np.uint8)) x = x.resize((int(shape[1] * c), int(shape[0] * c))) x = x.resize((shape[1], shape[0])) return np.asarray(x)
Conduct jpeg compression to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied jpeg compression. def jpeg_compression(x, severity=1): """Conduct jpeg compression to images. Args: x: numpy array, uncorrupted image, assumed to have uint8 pixel in [0,255]. severity: integer, severity of corruption. Returns: numpy array, image with uint8 pixels in [0,255]. Applied jpeg compression. """ c = [25, 18, 15, 10, 7][severity - 1] x = tfds.core.lazy_imports.PIL_Image.fromarray(x.astype(np.uint8)) output = io.BytesIO() x.save(output, 'JPEG', quality=c) output.seek(0) x = tfds.core.lazy_imports.PIL_Image.open(output) return np.asarray(x)
Temporarily assign obj.attr to value. def temporary_assignment(obj, attr, value): """Temporarily assign obj.attr to value.""" original = getattr(obj, attr, None) setattr(obj, attr, value) yield setattr(obj, attr, original)
Iterate over items of dictionaries grouped by their keys. def zip_dict(*dicts): """Iterate over items of dictionaries grouped by their keys.""" for key in set(itertools.chain(*dicts)): # set merge all keys # Will raise KeyError if the dict don't have the same keys yield key, tuple(d[key] for d in dicts)
Apply a function recursively to each element of a nested data struct. def map_nested(function, data_struct, dict_only=False, map_tuple=False): """Apply a function recursively to each element of a nested data struct.""" # Could add support for more exotic data_struct, like OrderedDict if isinstance(data_struct, dict): return { k: map_nested(function, v, dict_only, map_tuple) for k, v in data_struct.items() } elif not dict_only: types = [list] if map_tuple: types.append(tuple) if isinstance(data_struct, tuple(types)): mapped = [map_nested(function, v, dict_only, map_tuple) for v in data_struct] if isinstance(data_struct, list): return mapped else: return tuple(mapped) # Singleton return function(data_struct)
Zip data struct together and return a data struct with the same shape. def zip_nested(arg0, *args, **kwargs): """Zip data struct together and return a data struct with the same shape.""" # Python 2 do not support kwargs only arguments dict_only = kwargs.pop("dict_only", False) assert not kwargs # Could add support for more exotic data_struct, like OrderedDict if isinstance(arg0, dict): return { k: zip_nested(*a, dict_only=dict_only) for k, a in zip_dict(arg0, *args) } elif not dict_only: if isinstance(arg0, list): return [zip_nested(*a, dict_only=dict_only) for a in zip(arg0, *args)] # Singleton return (arg0,) + args
Simulate proto inheritance. By default, protobuf do not support direct inheritance, so this decorator simulates inheritance to the class to which it is applied. Example: ``` @as_proto_class(proto.MyProto) class A(object): def custom_method(self): return self.proto_field * 10 p = proto.MyProto(proto_field=123) a = A() a.CopyFrom(p) # a is like a proto object assert a.proto_field == 123 a.custom_method() # But has additional methods ``` Args: proto_cls: The protobuf class to inherit from Returns: decorated_cls: The decorated class def as_proto_cls(proto_cls): """Simulate proto inheritance. By default, protobuf do not support direct inheritance, so this decorator simulates inheritance to the class to which it is applied. Example: ``` @as_proto_class(proto.MyProto) class A(object): def custom_method(self): return self.proto_field * 10 p = proto.MyProto(proto_field=123) a = A() a.CopyFrom(p) # a is like a proto object assert a.proto_field == 123 a.custom_method() # But has additional methods ``` Args: proto_cls: The protobuf class to inherit from Returns: decorated_cls: The decorated class """ def decorator(cls): """Decorator applied to the class.""" class ProtoCls(object): """Base class simulating the protobuf.""" def __init__(self, *args, **kwargs): super(ProtoCls, self).__setattr__( "_ProtoCls__proto", proto_cls(*args, **kwargs), ) def __getattr__(self, attr_name): return getattr(self.__proto, attr_name) def __setattr__(self, attr_name, new_value): try: return setattr(self.__proto, attr_name, new_value) except AttributeError: return super(ProtoCls, self).__setattr__(attr_name, new_value) def __eq__(self, other): return self.__proto, other.get_proto() def get_proto(self): return self.__proto def __repr__(self): return "<{cls_name}\n{proto_repr}\n>".format( cls_name=cls.__name__, proto_repr=repr(self.__proto)) decorator_cls = type(cls.__name__, (cls, ProtoCls), { "__doc__": cls.__doc__, }) return decorator_cls return decorator
Path to tensorflow_datasets directory. def tfds_dir(): """Path to tensorflow_datasets directory.""" return os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
Writes to path atomically, by writing to temp file and renaming it. def atomic_write(path, mode): """Writes to path atomically, by writing to temp file and renaming it.""" tmp_path = "%s%s_%s" % (path, constants.INCOMPLETE_SUFFIX, uuid.uuid4().hex) with tf.io.gfile.GFile(tmp_path, mode) as file_: yield file_ tf.io.gfile.rename(tmp_path, path, overwrite=True)
Given a hash constructor, returns checksum digest and size of file. def read_checksum_digest(path, checksum_cls=hashlib.sha256): """Given a hash constructor, returns checksum digest and size of file.""" checksum = checksum_cls() size = 0 with tf.io.gfile.GFile(path, "rb") as f: while True: block = f.read(io.DEFAULT_BUFFER_SIZE) size += len(block) if not block: break checksum.update(block) return checksum.hexdigest(), size
Reraise an exception with an additional message. def reraise(additional_msg): """Reraise an exception with an additional message.""" exc_type, exc_value, exc_traceback = sys.exc_info() msg = str(exc_value) + "\n" + additional_msg six.reraise(exc_type, exc_type(msg), exc_traceback)
Get attr that handles dots in attr name. def rgetattr(obj, attr, *args): """Get attr that handles dots in attr name.""" def _getattr(obj, attr): return getattr(obj, attr, *args) return functools.reduce(_getattr, [obj] + attr.split("."))
Returns SplitGenerators. def _split_generators(self, dl_manager): """Returns SplitGenerators.""" image_tar_file = os.path.join(dl_manager.manual_dir, self.builder_config.file_name) if not tf.io.gfile.exists(image_tar_file): # The current celebahq generation code depends on a concrete version of # pillow library and cannot be easily ported into tfds. msg = "You must download the dataset files manually and place them in: " msg += dl_manager.manual_dir msg += " as .tar files. See testing/test_data/fake_examples/celeb_a_hq " raise AssertionError(msg) return [ tfds.core.SplitGenerator( name=tfds.Split.TRAIN, num_shards=50, gen_kwargs={"archive": dl_manager.iter_archive(image_tar_file)}, ) ]
This function returns the examples in the raw (text) form. def _generate_examples(self, source_file, target_file): """This function returns the examples in the raw (text) form.""" with tf.io.gfile.GFile(source_file) as f: source_sentences = f.read().split("\n") with tf.io.gfile.GFile(target_file) as f: target_sentences = f.read().split("\n") assert len(target_sentences) == len( source_sentences), "Sizes do not match: %d vs %d for %s vs %s." % (len( source_sentences), len(target_sentences), source_file, target_file) source, target = self.builder_config.language_pair for l1, l2 in zip(source_sentences, target_sentences): result = {source: l1, target: l2} # Make sure that both translations are non-empty. if all(result.values()): yield result
This function returns the examples in the raw (text) form. def _generate_examples(self, filepath): """This function returns the examples in the raw (text) form.""" rows_per_pair_id = collections.defaultdict(list) with tf.io.gfile.GFile(filepath) as f: reader = csv.DictReader(f, delimiter='\t', quoting=csv.QUOTE_NONE) for row in reader: rows_per_pair_id[row['pairID']].append(row) for rows in six.itervalues(rows_per_pair_id): premise = {row['language']: row['sentence1'] for row in rows} hypothesis = {row['language']: row['sentence2'] for row in rows} yield { 'premise': premise, 'hypothesis': hypothesis, 'label': rows[0]['gold_label'], }
Yields examples. def _generate_example(self, data_path, image_id): """Yields examples.""" image_filepath = os.path.join( data_path, "VOCdevkit/VOC2007/JPEGImages", "{}.jpg".format(image_id)) annon_filepath = os.path.join( data_path, "VOCdevkit/VOC2007/Annotations", "{}.xml".format(image_id)) def _get_example_objects(): """Function to get all the objects from the annotation XML file.""" with tf.io.gfile.GFile(annon_filepath, "r") as f: root = xml.etree.ElementTree.parse(f).getroot() size = root.find("size") width = float(size.find("width").text) height = float(size.find("height").text) for obj in root.findall("object"): # Get object's label name. label = obj.find("name").text.lower() # Get objects' pose name. pose = obj.find("pose").text.lower() is_truncated = (obj.find("truncated").text == "1") is_difficult = (obj.find("difficult").text == "1") bndbox = obj.find("bndbox") xmax = float(bndbox.find("xmax").text) xmin = float(bndbox.find("xmin").text) ymax = float(bndbox.find("ymax").text) ymin = float(bndbox.find("ymin").text) yield { "label": label, "pose": pose, "bbox": tfds.features.BBox( ymin / height, xmin / width, ymax / height, xmax / width), "is_truncated": is_truncated, "is_difficult": is_difficult, } objects = list(_get_example_objects()) # Use set() to remove duplicates labels = sorted(set(obj["label"] for obj in objects)) labels_no_difficult = sorted(set( obj["label"] for obj in objects if obj["is_difficult"] == 0 )) return { "image": image_filepath, "image/filename": image_id + ".jpg", "objects": objects, "labels": labels, "labels_no_difficult": labels_no_difficult, }
Update the encoding format. def set_encoding_format(self, encoding_format): """Update the encoding format.""" supported = ENCODE_FN.keys() if encoding_format not in supported: raise ValueError('`encoding_format` must be one of %s.' % supported) self._encoding_format = encoding_format
Update the shape. def set_shape(self, shape): """Update the shape.""" channels = shape[-1] acceptable_channels = ACCEPTABLE_CHANNELS[self._encoding_format] if channels not in acceptable_channels: raise ValueError('Acceptable `channels` for %s: %s (was %s)' % ( self._encoding_format, acceptable_channels, channels)) self._shape = tuple(shape)
Returns np_image encoded as jpeg or png. def _encode_image(self, np_image): """Returns np_image encoded as jpeg or png.""" if np_image.dtype != np.uint8: raise ValueError('Image should be uint8. Detected: %s.' % np_image.dtype) utils.assert_shape_match(np_image.shape, self._shape) return self._runner.run(ENCODE_FN[self._encoding_format], np_image)
Convert the given image into a dict convertible to tf example. def encode_example(self, image_or_path_or_fobj): """Convert the given image into a dict convertible to tf example.""" if isinstance(image_or_path_or_fobj, np.ndarray): encoded_image = self._encode_image(image_or_path_or_fobj) elif isinstance(image_or_path_or_fobj, six.string_types): with tf.io.gfile.GFile(image_or_path_or_fobj, 'rb') as image_f: encoded_image = image_f.read() else: encoded_image = image_or_path_or_fobj.read() return encoded_image
Reconstruct the image from the tf example. def decode_example(self, example): """Reconstruct the image from the tf example.""" img = tf.image.decode_image( example, channels=self._shape[-1], dtype=tf.uint8) img.set_shape(self._shape) return img
See base class for details. def save_metadata(self, data_dir, feature_name=None): """See base class for details.""" filepath = _get_metadata_filepath(data_dir, feature_name) with tf.io.gfile.GFile(filepath, 'w') as f: json.dump({ 'shape': [-1 if d is None else d for d in self._shape], 'encoding_format': self._encoding_format, }, f, sort_keys=True)
See base class for details. def load_metadata(self, data_dir, feature_name=None): """See base class for details.""" # Restore names if defined filepath = _get_metadata_filepath(data_dir, feature_name) if tf.io.gfile.exists(filepath): with tf.io.gfile.GFile(filepath, 'r') as f: info_data = json.load(f) self.set_encoding_format(info_data['encoding_format']) self.set_shape([None if d == -1 else d for d in info_data['shape']])
Create a moving image sequence from the given image a left padding values. Args: image: [in_h, in_w, n_channels] uint8 array pad_lefts: [sequence_length, 2] int32 array of left padding values total_padding: tensor of padding values, (pad_h, pad_w) Returns: [sequence_length, out_h, out_w, n_channels] uint8 image sequence, where out_h = in_h + pad_h, out_w = in_w + out_w def _create_moving_sequence(image, pad_lefts, total_padding): """Create a moving image sequence from the given image a left padding values. Args: image: [in_h, in_w, n_channels] uint8 array pad_lefts: [sequence_length, 2] int32 array of left padding values total_padding: tensor of padding values, (pad_h, pad_w) Returns: [sequence_length, out_h, out_w, n_channels] uint8 image sequence, where out_h = in_h + pad_h, out_w = in_w + out_w """ with tf.name_scope("moving_sequence"): def get_padded_image(args): pad_left, = args pad_right = total_padding - pad_left padding = tf.stack([pad_left, pad_right], axis=-1) z = tf.zeros((1, 2), dtype=pad_left.dtype) padding = tf.concat([padding, z], axis=0) return tf.pad(image, padding) padded_images = tf.map_fn( get_padded_image, [pad_lefts], dtype=tf.uint8, infer_shape=False, back_prop=False) return padded_images
Construct a linear trajectory from x0. Args: x0: N-D float tensor. velocity: N-D float tensor t: [sequence_length]-length float tensor Returns: x: [sequence_length, ndims] float tensor. def _get_linear_trajectory(x0, velocity, t): """Construct a linear trajectory from x0. Args: x0: N-D float tensor. velocity: N-D float tensor t: [sequence_length]-length float tensor Returns: x: [sequence_length, ndims] float tensor. """ x0 = tf.convert_to_tensor(x0) velocity = tf.convert_to_tensor(velocity) t = tf.convert_to_tensor(t) if x0.shape.ndims != 1: raise ValueError("x0 must be a rank 1 tensor") if velocity.shape.ndims != 1: raise ValueError("velocity must be a rank 1 tensor") if t.shape.ndims != 1: raise ValueError("t must be a rank 1 tensor") x0 = tf.expand_dims(x0, axis=0) velocity = tf.expand_dims(velocity, axis=0) dx = velocity * tf.expand_dims(t, axis=-1) linear_trajectories = x0 + dx assert linear_trajectories.shape.ndims == 2, \ "linear_trajectories should be a rank 2 tensor" return linear_trajectories
Turn simple static images into sequences of the originals bouncing around. Adapted from Srivastava et al. http://www.cs.toronto.edu/~nitish/unsupervised_video/ Example usage: ```python import tensorflow as tf import tensorflow_datasets as tfds from tensorflow_datasets.video import moving_sequence tf.compat.v1.enable_eager_execution() def animate(sequence): import numpy as np import matplotlib.pyplot as plt import matplotlib.animation as animation sequence = np.squeeze(sequence, axis=-1) fig = plt.figure() plt.axis("off") ims = [[plt.imshow(im, cmap="gray", animated=True)] for im in sequence] # don't remove `anim =` as linter may suggets # weird behaviour, plot will freeze on last frame anim = animation.ArtistAnimation( fig, ims, interval=50, blit=True, repeat_delay=100) plt.show() plt.close() tf.enable_eager_execution() mnist_ds = tfds.load("mnist", split=tfds.Split.TRAIN, as_supervised=True) mnist_ds = mnist_ds.repeat().shuffle(1024) def map_fn(image, label): sequence = moving_sequence.image_as_moving_sequence( image, sequence_length=20) return sequence.image_sequence moving_mnist_ds = mnist_ds.map(map_fn).batch(2).map( lambda x: dict(image_sequence=tf.reduce_max(x, axis=0))) # # for comparison with test data provided by original authors # moving_mnist_ds = tfds.load("moving_mnist", split=tfds.Split.TEST) for seq in moving_mnist_ds: animate(seq["image_sequence"].numpy()) ``` Args: image: [in_h, in_w, n_channels] tensor defining the sub-image to be bouncing around. sequence_length: int, length of sequence. output_size: (out_h, out_w) size returned images. velocity: scalar speed or 2D velocity of image. If scalar, the 2D velocity is randomly generated with this magnitude. This is the normalized distance moved each time step by the sub-image, where normalization occurs over the feasible distance the sub-image can move e.g if the input image is [10 x 10] and the output image is [60 x 60], a speed of 0.1 means the sub-image moves (60 - 10) * 0.1 = 5 pixels per time step. start_position: 2D float32 normalized initial position of each image in [0, 1]. Randomized uniformly if not given. Returns: `MovingSequence` namedtuple containing: `image_sequence`: [sequence_length, out_h, out_w, n_channels] image at each time step. padded values are all zero. Same dtype as input image. `trajectory`: [sequence_length, 2] float32 in [0, 1] 2D normalized coordinates of the image at every time step. `start_position`: 2D float32 initial position in [0, 1]. 2D normalized initial position of image. Same as input if provided, otherwise the randomly value generated. `velocity`: 2D float32 normalized velocity. Same as input velocity if provided as a 2D tensor, otherwise the random velocity generated. def image_as_moving_sequence( image, sequence_length=20, output_size=(64, 64), velocity=0.1, start_position=None): """Turn simple static images into sequences of the originals bouncing around. Adapted from Srivastava et al. http://www.cs.toronto.edu/~nitish/unsupervised_video/ Example usage: ```python import tensorflow as tf import tensorflow_datasets as tfds from tensorflow_datasets.video import moving_sequence tf.compat.v1.enable_eager_execution() def animate(sequence): import numpy as np import matplotlib.pyplot as plt import matplotlib.animation as animation sequence = np.squeeze(sequence, axis=-1) fig = plt.figure() plt.axis("off") ims = [[plt.imshow(im, cmap="gray", animated=True)] for im in sequence] # don't remove `anim =` as linter may suggets # weird behaviour, plot will freeze on last frame anim = animation.ArtistAnimation( fig, ims, interval=50, blit=True, repeat_delay=100) plt.show() plt.close() tf.enable_eager_execution() mnist_ds = tfds.load("mnist", split=tfds.Split.TRAIN, as_supervised=True) mnist_ds = mnist_ds.repeat().shuffle(1024) def map_fn(image, label): sequence = moving_sequence.image_as_moving_sequence( image, sequence_length=20) return sequence.image_sequence moving_mnist_ds = mnist_ds.map(map_fn).batch(2).map( lambda x: dict(image_sequence=tf.reduce_max(x, axis=0))) # # for comparison with test data provided by original authors # moving_mnist_ds = tfds.load("moving_mnist", split=tfds.Split.TEST) for seq in moving_mnist_ds: animate(seq["image_sequence"].numpy()) ``` Args: image: [in_h, in_w, n_channels] tensor defining the sub-image to be bouncing around. sequence_length: int, length of sequence. output_size: (out_h, out_w) size returned images. velocity: scalar speed or 2D velocity of image. If scalar, the 2D velocity is randomly generated with this magnitude. This is the normalized distance moved each time step by the sub-image, where normalization occurs over the feasible distance the sub-image can move e.g if the input image is [10 x 10] and the output image is [60 x 60], a speed of 0.1 means the sub-image moves (60 - 10) * 0.1 = 5 pixels per time step. start_position: 2D float32 normalized initial position of each image in [0, 1]. Randomized uniformly if not given. Returns: `MovingSequence` namedtuple containing: `image_sequence`: [sequence_length, out_h, out_w, n_channels] image at each time step. padded values are all zero. Same dtype as input image. `trajectory`: [sequence_length, 2] float32 in [0, 1] 2D normalized coordinates of the image at every time step. `start_position`: 2D float32 initial position in [0, 1]. 2D normalized initial position of image. Same as input if provided, otherwise the randomly value generated. `velocity`: 2D float32 normalized velocity. Same as input velocity if provided as a 2D tensor, otherwise the random velocity generated. """ ndims = 2 image = tf.convert_to_tensor(image) if image.shape.ndims != 3: raise ValueError("image must be rank 3, got %s" % str(image)) output_size = tf.TensorShape(output_size) if len(output_size) != ndims: raise ValueError("output_size must have exactly %d elements, got %s" % (ndims, output_size)) image_shape = tf.shape(image) if start_position is None: start_position = tf.random.uniform((ndims,), dtype=tf.float32) elif start_position.shape != (ndims,): raise ValueError("start_positions must (%d,)" % ndims) velocity = tf.convert_to_tensor(velocity, dtype=tf.float32) if velocity.shape.ndims == 0: velocity = _get_random_unit_vector(ndims, tf.float32) * velocity elif velocity.shape.ndims != 1: raise ValueError("velocity must be rank 0 or rank 1, got %s" % velocity) t = tf.range(sequence_length, dtype=tf.float32) trajectory = _get_linear_trajectory(start_position, velocity, t) trajectory = _bounce_to_bbox(trajectory) total_padding = output_size - image_shape[:2] if not tf.executing_eagerly(): cond = tf.compat.v1.assert_greater(total_padding, -1) with tf.control_dependencies([cond]): total_padding = tf.identity(total_padding) sequence_pad_lefts = tf.cast( tf.math.round(trajectory * tf.cast(total_padding, tf.float32)), tf.int32) sequence = _create_moving_sequence(image, sequence_pad_lefts, total_padding) sequence.set_shape( [sequence_length] + output_size.as_list() + [image.shape[-1]]) return MovingSequence( image_sequence=sequence, trajectory=trajectory, start_position=start_position, velocity=velocity)
Returns splits. def _split_generators(self, dl_manager): """Returns splits.""" dl_urls = { split: _BASE_DOWNLOAD_PATH + "%s.tfrecord" % split for split in _SPLITS } dl_urls["instrument_labels"] = (_BASE_DOWNLOAD_PATH + "instrument_labels.txt") dl_paths = dl_manager.download_and_extract(dl_urls) instrument_labels = tf.io.gfile.GFile(dl_paths["instrument_labels"], "r").read().strip().split("\n") self.info.features["instrument"]["label"].names = instrument_labels return [ tfds.core.SplitGenerator( # pylint: disable=g-complex-comprehension name=split, num_shards=_SPLIT_SHARDS[split], gen_kwargs={"path": dl_paths[split]}) for split in _SPLITS ]
Return the tuple (major, minor, patch) version extracted from the str. def _str_to_version(version_str, allow_wildcard=False): """Return the tuple (major, minor, patch) version extracted from the str.""" reg = _VERSION_WILDCARD_REG if allow_wildcard else _VERSION_RESOLVED_REG res = reg.match(version_str) if not res: msg = "Invalid version '{}'. Format should be x.y.z".format(version_str) if allow_wildcard: msg += " with {x,y,z} being digits or wildcard." else: msg += " with {x,y,z} being digits." raise ValueError(msg) return tuple( v if v == "*" else int(v) for v in [res.group("major"), res.group("minor"), res.group("patch")])
Returns True if other_version matches. Args: other_version: string, of the form "x[.y[.x]]" where {x,y,z} can be a number or a wildcard. def match(self, other_version): """Returns True if other_version matches. Args: other_version: string, of the form "x[.y[.x]]" where {x,y,z} can be a number or a wildcard. """ major, minor, patch = _str_to_version(other_version, allow_wildcard=True) return (major in [self.major, "*"] and minor in [self.minor, "*"] and patch in [self.patch, "*"])
Returns labels for validation. Args: val_path: path to TAR file containing validation images. It is used to retrieve the name of pictures and associate them to labels. Returns: dict, mapping from image name (str) to label (str). def _get_validation_labels(val_path): """Returns labels for validation. Args: val_path: path to TAR file containing validation images. It is used to retrieve the name of pictures and associate them to labels. Returns: dict, mapping from image name (str) to label (str). """ labels_path = tfds.core.get_tfds_path(_VALIDATION_LABELS_FNAME) with tf.io.gfile.GFile(labels_path) as labels_f: labels = labels_f.read().strip().split('\n') with tf.io.gfile.GFile(val_path, 'rb') as tar_f_obj: tar = tarfile.open(mode='r:', fileobj=tar_f_obj) images = sorted(tar.getnames()) return dict(zip(images, labels))
Yields examples. def _generate_examples(self, archive, validation_labels=None): """Yields examples.""" if validation_labels: # Validation split for example in self._generate_examples_validation(archive, validation_labels): yield example # Training split. Main archive contains archives names after a synset noun. # Each sub-archive contains pictures associated to that synset. for fname, fobj in archive: label = fname[:-4] # fname is something like 'n01632458.tar' # TODO(b/117643231): in py3, the following lines trigger tarfile module # to call `fobj.seekable()`, which Gfile doesn't have. We should find an # alternative, as this loads ~150MB in RAM. fobj_mem = io.BytesIO(fobj.read()) for image_fname, image_fobj in tfds.download.iter_archive( fobj_mem, tfds.download.ExtractMethod.TAR): yield { 'file_name': image_fname, 'image': image_fobj, 'label': label, }
Whether any of the filenames exist. def do_files_exist(filenames): """Whether any of the filenames exist.""" preexisting = [tf.io.gfile.exists(f) for f in filenames] return any(preexisting)
Returns a temporary filename based on filename. def get_incomplete_path(filename): """Returns a temporary filename based on filename.""" random_suffix = "".join( random.choice(string.ascii_uppercase + string.digits) for _ in range(6)) return filename + ".incomplete" + random_suffix