← Search

Text · dataset · 2017

SPEECH-COCO

Listed in DataCite

SPEECH-COCO is an augmentation of MS-COCO dataset where speech is added to image and text.

Description

Speech captions were generated using text-to-speech (TTS) synthesis resulting in 616,767 spoken captions (>600h) paired with images. Disfluencies and speed perturbation were added to the signal in order to sound more natural.

Each speech signal (WAV) is paired with a JSON file containing exact timecode for each word/syllable/phoneme in the spoken caption. Such a corpus could be used for Language and Vision (LaVi) tasks including speech input or output instead of text.

Links

Topics

Stated by source
Languages and literature
Inferred from text
Audio 65% · Image 75% · Text 75%
Provenance · 1 source records, 12 field assertions
SourceKeyLast seenRaw
DataCite10.18709/perscido.2017.06.ds8015 d agoJSON v1
FieldAssertionExtractorEvidence
access_levelsource · DataCiteconnector:datacite@1.0.0/data/attributes/rightsList
byte_sizesource · DataCiteconnector:datacite@1.0.0
concepts[field].fos:languages-and-literaturesource · DataCiteconnector:datacite@1.0.0
concepts[field].local:field:computer-science-aimapping · DataCitevocabulary-mapper@1.0.0keywords['Computer Science']
concepts[modality].local:modality:audioenrichment · DataCitekeyword-concept-rules@1.0.0title+description (65%)
concepts[modality].local:modality:imageenrichment · DataCitekeyword-concept-rules@1.0.0title+description (75%)
concepts[modality].local:modality:textenrichment · DataCitekeyword-concept-rules@1.0.0title+description (75%)
descriptionsource · DataCiteconnector:datacite@1.0.0/data/attributes/descriptions
licensesource · DataCiteconnector:datacite@1.0.0/data/attributes/rightsList
publication_datesource · DataCiteconnector:datacite@1.0.0/data/attributes/dates
titlesource · DataCiteconnector:datacite@1.0.0/data/attributes/titles/0/title
version_labelsource · DataCiteconnector:datacite@1.0.0