A Sparse Transformer-Based Approach for Image Captioning
Image Captioning is the task of providing a natural language description for an image. It has caught significant amounts of attention from both computer vision and natural language processing communities. Most image captioning models adopt deep encoder-decoder architectures to achieve state-of-the-a...
Guardado en:
Autores principales: | , , , , |
---|---|
Formato: | article |
Lenguaje: | EN |
Publicado: |
IEEE
2020
|
Materias: | |
Acceso en línea: | https://doaj.org/article/8d37acadce0441f6b826f861c201713c |
Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
id |
oai:doaj.org-article:8d37acadce0441f6b826f861c201713c |
---|---|
record_format |
dspace |
spelling |
oai:doaj.org-article:8d37acadce0441f6b826f861c201713c2021-11-19T00:05:19ZA Sparse Transformer-Based Approach for Image Captioning2169-353610.1109/ACCESS.2020.3024639https://doaj.org/article/8d37acadce0441f6b826f861c201713c2020-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9199872/https://doaj.org/toc/2169-3536Image Captioning is the task of providing a natural language description for an image. It has caught significant amounts of attention from both computer vision and natural language processing communities. Most image captioning models adopt deep encoder-decoder architectures to achieve state-of-the-art performances. However, it is difficult to model knowledge on relationships between input image region pairs in the encoder. Furthermore, the word in the decoder hardly knows the correlation to specific image regions. In this article, a novel deep encoder-decoder model is proposed for image captioning which is developed on sparse Transformer framework. The encoder adopts a multi-level representation of image features based on self-attention to exploit low-level and high-level features, naturally the correlations between image region pairs are adequately modeled as self-attention operation can be seen as a way of encoding pairwise relationships. The decoder improves the concentration of multi-head self-attention on the global context by explicitly selecting the most relevant segments at each row of the attention matrix. It can help the model focus on the more contributing image regions and generate more accurate words in the context. Experiments demonstrate that our model outperforms previous methods and achieves higher performance on MSCOCO and Flickr30k datasets. Our code is available at <uri>https://github.com/2014gaokao/ImageCaptioning</uri>.Zhou LeiCongcong ZhouShengbo ChenYiyong HuangXianrui LiuIEEEarticleImage captioningself-attentionexplict sparselocal adaptive thresholdElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 8, Pp 213437-213446 (2020) |
institution |
DOAJ |
collection |
DOAJ |
language |
EN |
topic |
Image captioning self-attention explict sparse local adaptive threshold Electrical engineering. Electronics. Nuclear engineering TK1-9971 |
spellingShingle |
Image captioning self-attention explict sparse local adaptive threshold Electrical engineering. Electronics. Nuclear engineering TK1-9971 Zhou Lei Congcong Zhou Shengbo Chen Yiyong Huang Xianrui Liu A Sparse Transformer-Based Approach for Image Captioning |
description |
Image Captioning is the task of providing a natural language description for an image. It has caught significant amounts of attention from both computer vision and natural language processing communities. Most image captioning models adopt deep encoder-decoder architectures to achieve state-of-the-art performances. However, it is difficult to model knowledge on relationships between input image region pairs in the encoder. Furthermore, the word in the decoder hardly knows the correlation to specific image regions. In this article, a novel deep encoder-decoder model is proposed for image captioning which is developed on sparse Transformer framework. The encoder adopts a multi-level representation of image features based on self-attention to exploit low-level and high-level features, naturally the correlations between image region pairs are adequately modeled as self-attention operation can be seen as a way of encoding pairwise relationships. The decoder improves the concentration of multi-head self-attention on the global context by explicitly selecting the most relevant segments at each row of the attention matrix. It can help the model focus on the more contributing image regions and generate more accurate words in the context. Experiments demonstrate that our model outperforms previous methods and achieves higher performance on MSCOCO and Flickr30k datasets. Our code is available at <uri>https://github.com/2014gaokao/ImageCaptioning</uri>. |
format |
article |
author |
Zhou Lei Congcong Zhou Shengbo Chen Yiyong Huang Xianrui Liu |
author_facet |
Zhou Lei Congcong Zhou Shengbo Chen Yiyong Huang Xianrui Liu |
author_sort |
Zhou Lei |
title |
A Sparse Transformer-Based Approach for Image Captioning |
title_short |
A Sparse Transformer-Based Approach for Image Captioning |
title_full |
A Sparse Transformer-Based Approach for Image Captioning |
title_fullStr |
A Sparse Transformer-Based Approach for Image Captioning |
title_full_unstemmed |
A Sparse Transformer-Based Approach for Image Captioning |
title_sort |
sparse transformer-based approach for image captioning |
publisher |
IEEE |
publishDate |
2020 |
url |
https://doaj.org/article/8d37acadce0441f6b826f861c201713c |
work_keys_str_mv |
AT zhoulei asparsetransformerbasedapproachforimagecaptioning AT congcongzhou asparsetransformerbasedapproachforimagecaptioning AT shengbochen asparsetransformerbasedapproachforimagecaptioning AT yiyonghuang asparsetransformerbasedapproachforimagecaptioning AT xianruiliu asparsetransformerbasedapproachforimagecaptioning AT zhoulei sparsetransformerbasedapproachforimagecaptioning AT congcongzhou sparsetransformerbasedapproachforimagecaptioning AT shengbochen sparsetransformerbasedapproachforimagecaptioning AT yiyonghuang sparsetransformerbasedapproachforimagecaptioning AT xianruiliu sparsetransformerbasedapproachforimagecaptioning |
_version_ |
1718420678586138624 |