AVAtt:面向艺术绘画视觉注意力的自由注视眼动追踪数据集与基准
英文标题:AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings
针对现有显著性与视线轨迹模型几乎仅基于自然摄影图像构建的问题,研究团队推出了面向艺术绘画的自由注视眼动追踪数据集与基准 AVAtt。该数据集涵盖 16 个艺术流派、跨越 5 个世纪的画作,提供了细粒度注视轨迹、眼屏距离、显著图及丰富的艺术元数据。基准测试评估了经典与深度显著性模型,并对比了 44 种轨迹相似度评估指标(包括基于 CNN、ViT 和自监督编码器的新型指标族)。分析证实面部具有极高且迅速(0.5秒内)的视线吸引力,艺术风格与构图对观者视觉注意有显著塑造作用,为开发具备艺术感知能力的计算机视觉模型提供了坚实基础。
672 B
612 B
380 B
24.2 KB
34.7 KB
中文
[提交于 2026 年 10 月 21 日]
标题:AVAtt:面向艺术绘画视觉注意力的自由注视眼动追踪数据集与基准(AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings)
阅读由 Mohamed Amine Kerkouri 及其他 3 位作者共同撰写的论文 PDF:《AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings》
摘要:显著性(Saliency)和视线注视轨迹(Scanpath)模型几乎完全是基于自然摄影图像开发建立的,然而绘画作品是由艺术家精心构图以引导观者视线的,且在纹理、色彩与抽象表达上均与摄影照片截然不同。为此,我们推出了 \avatt,一个涵盖 16 个艺术流派、跨越 5 个世纪绘画作品的自由注视眼动追踪数据集,包含来自 \nParticipants{} 位受试者对 \nImages{} 幅画作产生的 \nScanpaths{} 条注视轨迹与 \nFixations{} 次注视点记录(每幅画作有 \spPerImageMin--\spPerImageMax{} 名观察者)。针对每幅画作,我们发布了包含注视时序和眼屏距离的单人注视轨迹、注视点图(Fixation maps)、在 1 度视角下平滑处理的显著图(Saliency maps)、时间显著性视频以及丰富的艺术元数据。我们将经典显著性模型和深度显著性模型与人类观察者的一致性进行了基准对比,结果表明艺术感知型模型仍有巨大的提升空间;同时我们对 44 种注视轨迹相似度度量标准进行了排名评估,其中包括基于 10 种 CNN、ViT 及自监督编码器构建的新型无语言(language-free)图像编码器度量族。在消除观看偏差的受控分析下,我们表明:观者注视面部的频率比预期高出 \faceEnrich 倍且在半秒内就会发生;局部结构会吸引视线;根据图像内容拟合的构图规则能够预测人们注视的位置;并且艺术风格直接塑造了观者视线的一致程度。数据集、代码及 Python 软件包已开源发布于:此链接。
提交历史
来自:Mohamed Amine Kerkouri [
]
[v1] 2026 年 10 月 21 日 星期三 22:02:01 UTC (25,997 KB)
English / 原文
[Submitted on 21 Oct 2026]
Title:AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings
View a PDF of the paper titled AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings, by Mohamed Amine Kerkouri and 3 other authors
Abstract:Saliency and scanpath models are developed almost exclusively on natural photographs, yet paintings are composed by artists to guide the viewer's eye and differ from photographs in texture, colour and abstraction. We introduce \avatt, a free-viewing eye-tracking dataset on \nImages{} paintings spanning sixteen art movements and five centuries, with \nScanpaths{} scanpaths and \nFixations{} fixations from \nParticipants{} participants (\spPerImageMin--\spPerImageMax{} observers per painting). For every painting we release individual scanpaths with fixation timing and eye--screen distance, fixation maps, saliency maps smoothed at one degree of visual angle, temporal saliency videos and rich art metadata. We benchmark classic and deep saliency models against human agreement, revealing substantial headroom for art-aware models, and rank 44 scanpath-similarity measures, including a new language-free family of image-encoder metrics built on ten CNN, ViT and self-supervised encoders. With controls that remove viewing biases, we show that faces are fixated \faceEnrich$\times$ more than expected and within half a second, that local structure draws the eye, that composition rules fitted from image content predict where people look, and that art style shapes how much viewers agree. Dataset, code and a Python package are available at: this https URL.
Submission history
From: Mohamed Amine Kerkouri [
]
[v1]
Wed, 21 Oct 2026 22:02:01 UTC (25,997 KB)