# coding=utf-8 # Copyright 2020 The TensorFlow Datasets Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # Lint as: python3 """UCF-101 dataset from https://www.crcv.ucf.edu/data/UCF101.php.""" from __future__ import absolute_import from __future__ import division from __future__ import print_function import os from absl import logging import tensorflow.compat.v2 as tf import tensorflow_datasets.public_api as tfds UCF_101_URL = 'https://storage.googleapis.com/thumos14_files/UCF101_videos.zip' SPLITS_URL = ('https://www.crcv.ucf.edu/data/UCF101/' 'UCF101TrainTestSplits-RecognitionTask.zip') _CITATION = """\ @article{DBLP:journals/corr/abs-1212-0402, author = {Khurram Soomro and Amir Roshan Zamir and Mubarak Shah}, title = {{UCF101:} {A} Dataset of 101 Human Actions Classes From Videos in The Wild}, journal = {CoRR}, volume = {abs/1212.0402}, year = {2012}, url = {http://arxiv.org/abs/1212.0402}, archivePrefix = {arXiv}, eprint = {1212.0402}, timestamp = {Mon, 13 Aug 2018 16:47:45 +0200}, biburl = {https://dblp.org/rec/bib/journals/corr/abs-1212-0402}, bibsource = {dblp computer science bibliography, https://dblp.org} } """ _LABELS_FNAME = 'video/ucf101_labels.txt' class Ucf101Config(tfds.core.BuilderConfig): """"Configuration for UCF101 split and possible video rescaling.""" @tfds.core.disallow_positional_args def __init__(self, split_number, width=None, height=None, **kwargs): """The parameters specifying how the dataset will be processed. The dataset comes with three separate splits. You can specify which split you want in `split_number`. If `width` and `height` are set, the videos will be rescaled to have those heights and widths (using ffmpeg). Args: split_number: The split number, one of (1, 2, 3) width: An integer with the width or None. height: An integer with the height or None. **kwargs: Passed on to the constructor of `BuilderConfig`. """ super(Ucf101Config, self).__init__(**kwargs) if (width is None) ^ (height is None): raise ValueError('Either both dimensions should be set, or none of them') self.width = width self.height = height if split_number not in (1, 2, 3): raise ValueError('Unknown split number {}, should be 1, 2 or 3'.format( split_number)) self.split_number = split_number _VERSION = tfds.core.Version( '2.0.0', 'New split API (https://tensorflow.org/datasets/splits)') class Ucf101(tfds.core.GeneratorBasedBuilder): """Ucf101 action recognition dataset. Note that in contrast to the labels provided in the original dataset, here the labels start at zero, not at one. """ BUILDER_CONFIGS = [ Ucf101Config( name='ucf101_1_256', description='256x256 UCF with the first action recognition split.', width=256, height=256, split_number=1, version=_VERSION, ), Ucf101Config( name='ucf101_1', description='UCF with the action recognition split #1.', width=None, height=None, split_number=1, version=_VERSION, ), Ucf101Config( name='ucf101_2', description='UCF with the action recognition split #2.', width=None, height=None, split_number=2, version=_VERSION, ), Ucf101Config( name='ucf101_3', description='UCF with the action recognition split #3.', width=None, height=None, split_number=3, version=_VERSION, ), ] def _info(self): if self.builder_config.width is not None: if self.builder_config.height is None: raise ValueError('Provide either both height and width or none.') ffmpeg_extra_args = ( '-vf', 'scale={}x{}'.format(self.builder_config.height, self.builder_config.width)) else: ffmpeg_extra_args = [] video_shape = ( None, self.builder_config.height, self.builder_config.width, 3) labels_names_file = tfds.core.get_tfds_path(_LABELS_FNAME) features = tfds.features.FeaturesDict({ 'video': tfds.features.Video(video_shape, ffmpeg_extra_args=ffmpeg_extra_args, encoding_format='jpeg'), 'label': tfds.features.ClassLabel(names_file=labels_names_file), }) return tfds.core.DatasetInfo( builder=self, description='A 101-label video classification dataset.', features=features, homepage='https://www.crcv.ucf.edu/data-sets/ucf101/', citation=_CITATION, ) def _split_generators(self, dl_manager): splits_folder = 'ucfTrainTestlist' urls_to_download = { 'videos': UCF_101_URL, 'splits': SPLITS_URL, } downloaded_urls = dl_manager.download_and_extract(urls_to_download) return [ tfds.core.SplitGenerator( name=tfds.Split.TRAIN, gen_kwargs={ 'videos_dir': downloaded_urls['videos'], 'splits_dir': downloaded_urls['splits'], 'data_list': '{}/trainlist{:02d}.txt'.format( splits_folder, self.builder_config.split_number), }), tfds.core.SplitGenerator( name=tfds.Split.TEST, gen_kwargs={ 'videos_dir': downloaded_urls['videos'], 'splits_dir': downloaded_urls['splits'], 'data_list': '{}/testlist{:02d}.txt'.format( splits_folder, self.builder_config.split_number), }), ] def _generate_examples(self, videos_dir, splits_dir, data_list): data_list_path_path = os.path.join(splits_dir, data_list) with tf.io.gfile.GFile(data_list_path_path, 'r') as data_list_file: labels_and_paths = data_list_file.readlines() for label_and_path in sorted(labels_and_paths): # The train splits contain not only the filename, but also a digit # encoding the label separated by a space, which we ignore. label_and_path = label_and_path.strip().split(' ')[0] label, path = os.path.split(label_and_path) # Fix an inconsistency between the names in the list and in the zip file. path = path.replace('HandStandPushups', 'HandstandPushups') video_path = os.path.join(videos_dir, 'UCF101', path) if not tf.io.gfile.exists(video_path): logging.error('Example %s not found', video_path) continue # We extract the label from the filename. yield path, {'video': video_path, 'label': label}