Note: The datasets documented here are from HEAD and so not all are available
in the current tensorflow-datasets package. They are all accessible in our
nightly package tfds-nightly.
# Build the `tf.data.Dataset` pipeline.
ds, info = tfds.load('cifar10', split='train', shuffle_files=True, with_info=True)
ds = ds.shuffle(info.splits['train'].num_examples)
ds = ds.batch(32)
# `tfds.as_numpy` converts `tf.Tensor` -> `np.array`
for ex in tfds.as_numpy(ds):
# `int2str` returns the human readable label ('dog', 'car',...)
print(info.features['label'].int2str(ex['label']))AudioImageabstract_reasoningaflw2k3darcbeansbigearthnetbinarized_mnistbinary_alpha_digitscaltech101caltech_birds2010caltech_birds2011cars196cassavacats_vs_dogsceleb_aceleb_a_hqcifar10cifar100cifar10_1cifar10_corruptedcitrus_leavescityscapesclevrcmaterdbcoil100colorectal_histologycolorectal_histology_largecurated_breast_imaging_ddsmcycle_gandeep_weedsdiabetic_retinopathy_detectiondiv2kdmlabdownsampled_imagenetdspritesdtdduke_ultrasoundemnisteurosatfashion_mnistflicfood101geirhos_conflict_stimulihorses_or_humansi_naturalist2017image_label_folderimagenet2012imagenet2012_corruptedimagenet2012_subsetimagenet_resizedimagenetteimagewangkmnistlfwlost_and_foundlsunmalariamnistmnist_corruptedomniglotoxford_flowers102oxford_iiit_petpatch_camelyonpet_finderplaces365_smallplant_leavesplant_villageplantae_kquickdraw_bitmapresisc45rock_paper_scissorsscene_parse150shapes3dsmallnorbso2satstanford_dogsstanford_online_productssun397svhn_croppedtf_flowersthe300w_lpuc_mercedvgg_face2visual_domain_decathlon
Object_detectionStructuredSummarizationTextblimpc4cfqcivil_commentscos_edefinite_pronoun_resolutioneraser_multi_rcesnligapglueimdb_reviewslibrispeech_lmlm1bmath_datasetmovie_rationalesmulti_nlimulti_nli_mismatchnatural_questionsqa4mrescanscicitesnlisquadsuper_gluetiny_shakespearetrivia_qawiki40bwikipediaxnliyelp_polarity_reviews
TranslateVideo