Summary of Object-aware Query Perturbation For Cross-modal Image-text Retrieval, by Naoya Sogi et al.
Object-Aware Query Perturbation for Cross-Modal Image-Text Retrieval
by Naoya Sogi, Takashi Shibata, Makoto Terao
First submitted to arxiv on: 17 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Information Retrieval (cs.IR); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel framework for cross-modal image-text retrieval that leverages object-aware query perturbation. The authors recognize the limitations of pre-trained vision and language (V&L) models, which struggle with small objects due to rough alignment between words and images. In contrast, human cognition is object-centric, focusing on important objects even if they are small. To bridge this gap, the proposed method generates a key feature subspace from detected objects and perturbs queries using this subspace. This approach maintains the rich expressive power and retrieval performance of existing V&L models without additional fine-tuning. The authors demonstrate their method’s effectiveness on four public datasets, outperforming conventional algorithms. The comprehensive experiments showcase the benefits of incorporating object awareness into cross-modal image-text retrieval. The proposed framework is publicly available at this URL: https://github.com/NEC-N-SOGI/query-perturbation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to improve how computers find matching images and words on the internet. Right now, these computers struggle when the important objects in an image are small. Humans don’t have this problem because we focus on what’s most important, even if it’s tiny. The authors created a new method that helps computers be more like humans by paying attention to the important things in an image. This makes searching for images and words much better. The authors tested their method on many different datasets and showed that it works really well. You can find all the code they used online at this URL: https://github.com/NEC-N-SOGI/query-perturbation. |
Keywords
» Artificial intelligence » Alignment » Attention » Fine tuning
