arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2202.04007cs.CVcs.LG
收藏

2021年图像相似性挑战赛的结果和发现

Results and findings of the 2021 Image Similarity Challenge

Meta AI;Czech Technical University in Prague(布拉格捷克技术大学);DeNA;Baidu Research(百度研究院)
Zoë Papakipos, Giorgos Tolias, Tomas Jenicek, Ed Pizzi, Shuhei Yokoo, Wenhao Wang, Yifan Sun, Weipu Zhang, Yi Yang, Sanjay Addicam, Sergio Manuel Papadakis, Cri… 展开作者

Zoë Papakipos, Giorgos Tolias, Tomas Jenicek, Ed Pizzi, Shuhei Yokoo, Wenhao Wang, Yifan Sun, Weipu Zhang, Yi Yang, Sanjay Addicam, Sergio Manuel Papadakis, Cristian Canton Ferrer, Ondrej Chum, Matthijs Douze

摘要:

2021年图像相似性挑战赛引入了一个数据集,作为评估最新图像复制检测方法的新基准。有200人参加了比赛。本文对排名靠前的论文进行了定量和定性分析。看来,最困难的图像变换涉及严重的图像裁剪或隐藏到不相关的图像,结合局部像素扰动。获奖作品中的关键算法元素包括:强增强训练、自监督学习、分数归一化、显式重叠检测和全局描述符匹配,然后进行成对图像比较。

英文摘要:

The 2021 Image Similarity Challenge introduced a dataset to serve as a new benchmark to evaluate recent image copy detection methods. There were 200 participants to the competition. This paper presents a quantitative and qualitative analysis of the top submissions. It appears that the most difficult image transformations involve either severe image crops or hiding into unrelated images, combined with local pixel perturbations. The key algorithmic elements in the winning submissions are: training on strong augmentations, self-supervised learning, score normalization, explicit overlay detection, and global descriptor matching followed by pairwise image comparison.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑