Papers › VIVO: Visual Vocabulary Pre-Training for Novel Object Captioning

VIVO: Visual Vocabulary Pre-Training for Novel Object Captioning

28 Sep 2020arXiv:2009.13682archive 2025-07-28

Xiaowei Hu, Xi Yin, Kevin Lin, Lijuan Wang, Lei Zhang, Jianfeng Gao, Zicheng Liu

It is highly desirable yet challenging to generate image captions that can describe novel objects which are unseen in caption-labeled training data, a capability that is evaluated in the novel object captioning challenge (nocaps). In this challenge, no additional image-caption training data, other thanCOCO Captions, is allowed for model training. Thus, conventional Vision-Language Pre-training (VLP) methods cannot be applied. This paper presents VIsual VOcabulary pretraining (VIVO) that performs pre-training in the absence of caption annotations. By breaking the dependency of paired image-caption training data in VLP, VIVO can leverage large amounts of paired image-tag data to learn a visual vocabulary. This is done by pre-training a multi-layer Transformer model that learns to align image-level tags with their corresponding image region features. To address the unordered nature of image tags, VIVO uses a Hungarian matching loss with masked tag prediction to conduct pre-training. We validate the effectiveness of VIVO by fine-tuning the pre-trained model for image captioning. In addition, we perform an analysis of the visual-text alignment inferred by our model. The results show that our model can not only generate fluent image captions that describe novel objects, but also identify the locations of these objects. Our single model has achieved new state-of-the-art results on nocaps and surpassed the human CIDEr score.

PaperPDF

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image CaptioningObjectTAG

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Captioning nocaps entire Microsoft Cognitive Services team B1 85.62 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team B2 71.36 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team B3 53.62 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team B4 34.65 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team CIDEr 114.25 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team METEOR 31.27 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team ROUGE-L 61.2 #4 of 39 Archive leaderboard report
Image Captioning nocaps entire Microsoft Cognitive Services team SPICE 14.85 #4 of 39 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team B1 86.33 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team B2 72.83 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team B3 55.94 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team B4 37.97 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team CIDEr 112.82 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team METEOR 32.7 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team ROUGE-L 62.48 #6 of 41 Archive leaderboard report
Image Captioning nocaps in-domain Microsoft Cognitive Services team SPICE 15.22 #6 of 41 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team B1 86.48 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team B2 72.6 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team B3 55.26 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team B4 36.31 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team CIDEr 115.54 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team METEOR 31.8 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team ROUGE-L 61.9 #5 of 40 Archive leaderboard report
Image Captioning nocaps near-domain Microsoft Cognitive Services team SPICE 15.06 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team B1 81.73 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team B2 65.48 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team B3 45.58 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team B4 25.78 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team CIDEr 110.14 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team METEOR 28.17 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team ROUGE-L 57.57 #5 of 40 Archive leaderboard report
Image Captioning nocaps out-of-domain Microsoft Cognitive Services team SPICE 13.74 #5 of 40 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team B1 82.27 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team B2 66.04 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team B3 47.48 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team B4 28.95 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team CIDEr 100.12 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team METEOR 29.47 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team ROUGE-L 58.26 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD entire Microsoft Cognitive Services team SPICE 14.04 #5 of 12 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team B1 82.94 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team B2 67.56 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team B3 49.66 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team B4 32.07 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team CIDEr 100.62 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team METEOR 30.62 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team ROUGE-L 59.43 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD in-domain Microsoft Cognitive Services team SPICE 14.7 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team B1 82.88 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team B2 67.01 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team B3 48.73 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team B4 30.21 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team CIDEr 101.2 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team METEOR 30.0 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team ROUGE-L 58.76 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD near-domain Microsoft Cognitive Services team SPICE 14.27 #4 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team B1 79.44 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team B2 61.15 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team B3 41.03 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team B4 21.79 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team CIDEr 95.5 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team METEOR 26.56 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team ROUGE-L 55.49 #3 of 11 Archive leaderboard report
Image Captioning nocaps-XD out-of-domain Microsoft Cognitive Services team SPICE 12.66 #3 of 11 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Absolute Position EncodingsAdamAttentionBPEDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections