Corpus ID: 220363571

Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation

@article{Wang2020ImprovingWS,
  title={Improving Weakly Supervised Visual Grounding by Contrastive Knowledge Distillation},
  author={L. Wang and Jui-Ting Huang and Yanchao Li and Kun Xu and Zhengyuan Yang and Dong Yu},
  journal={ArXiv},
  year={2020},
  volume={abs/2007.01951}
}
  • L. Wang, Jui-Ting Huang, +3 authors Dong Yu
  • Published 2020
  • Computer Science
  • ArXiv
  • Weakly supervised phrase grounding aims at learning region-phrase correspondences using only image-sentence pairs. A major challenge thus lies in the missing links between image regions and sentence phrases during training. To address this challenge, we leverage a generic object detector at training time, and propose a contrastive learning framework that accounts for both region-phrase and imagesentence matching. Our core innovation is the learning of a region-phrase score function, based on… CONTINUE READING

    Figures, Tables, and Topics from this paper.

    Explore Further: Topics Discussed in This Paper

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 48 REFERENCES

    Modularized Textual Grounding for Counterfactual Resilience

    VIEW 11 EXCERPTS
    HIGHLY INFLUENTIAL

    Contrastive Learning for Weakly Supervised Phrase Grounding

    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL

    Knowledge Aided Consistency for Weakly Supervised Phrase Grounding

    VIEW 7 EXCERPTS
    HIGHLY INFLUENTIAL

    Phrase Localization Without Paired Training Examples

    VIEW 10 EXCERPTS
    HIGHLY INFLUENTIAL

    Unsupervised Textual Grounding: Linking Words to Image Concepts

    VIEW 8 EXCERPTS
    HIGHLY INFLUENTIAL

    WordNet: a lexical database for English

    VIEW 5 EXCERPTS
    HIGHLY INFLUENTIAL

    Adam: A Method for Stochastic Optimization

    VIEW 2 EXCERPTS
    HIGHLY INFLUENTIAL

    Flickr30k Entities: Collecting Region-to-Phrase Correspondences for Richer Image-to-Sentence Models

    VIEW 6 EXCERPTS
    HIGHLY INFLUENTIAL