Context-empowered Visual Attention Prediction in Pedestrian Scenarios

被引:1
作者
Vozniak, Igor [1 ]
Mueller, Philipp [1 ]
Hell, Lorena [1 ]
Lipp, Nils [1 ]
Abouelazm, Ahmed [1 ]
Mueller, Christian [1 ]
机构
[1] German Res Ctr Artificial Intelligence DFKI, Saarbrucken, Germany
来源
2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV) | 2023年
关键词
EYE-MOVEMENTS; SALIENCY; GAZE; BEHAVIORS; NETWORK;
D O I
10.1109/WACV56688.2023.00101
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Effective and flexible allocation of visual attention is key for pedestrians who have to navigate to a desired goal under different conditions of urgency and safety preferences. While automatic modelling of pedestrian attention holds great promise to improve simulations of pedestrian behavior, current saliency prediction approaches mostly focus on generic free-viewing scenarios and do not reflect the specific challenges present in pedestrian attention prediction. In this paper, we present Context-SalNET, a novel encoder-decoder architecture that explicitly addresses three key challenges of visual attention prediction in pedestrians: First, Context-SalNET explicitly models the context factors urgency and safety preference in the latent space of the encoder-decoder model. Second, we propose the exponentially weighted mean squared error loss (ew-MSE) that is able to better cope with the fact that only a small part of the ground truth saliency maps consist of non-zero entries. Third, we explicitly model epistemic uncertainty to account for the fact that training data for pedestrian attention prediction is limited. To evaluate Context-SalNET, we recorded the first dataset of pedestrian visual attention in VR that includes explicit variation of the context factors urgency and safety preference. Context-SalNET achieves clear improvements over state-of-the-art saliency prediction approaches as well as over ablations. Our novel dataset will be made fully available and can serve as a valuable resource for further research on pedestrian attention prediction.
引用
收藏
页码:950 / 960
页数:11
相关论文
共 76 条
  • [1] [Anonymous], 2015, PROC CVPR IEEE
  • [2] [Anonymous], 2018, P 32 INT C NEUR INF
  • [3] [Anonymous], 2020, EUR C COMP VIS
  • [4] [Anonymous], 2021, 2021 17 INT C MACH V
  • [5] [Anonymous], 2023, P IEEE C COMP VIS PA, DOI DOI 10.1080/23249935.2022.2033348
  • [6] HAIL: Modular Agent-Based Pedestrian Imitation Learning
    Antakli, Andre
    Vozniak, Igor
    Lipp, Nils
    Klusch, Matthias
    Muller, Christian
    [J]. ADVANCES IN PRACTICAL APPLICATIONS OF AGENTS, MULTI-AGENT SYSTEMS, AND SOCIAL GOOD: THE PAAMS COLLECTION, PAAMS 2021, 2021, 12946 : 27 - 39
  • [7] Bazilinskyy Pavlo, 2021, Advances in Human Aspects of Transportation. Proceedings of the AHFE 2021 Virtual Conference on Human Aspects of Transportation. Lecture Notes in Networks and Systems (LNNS 270), P147, DOI 10.1007/978-3-030-80012-3_18
  • [8] Saliency Prediction in the Deep Learning Era: Successes and Limitations
    Borji, Ali
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2021, 43 (02) : 679 - 700
  • [9] Defending Yarbus: Eye movements reveal observers' task
    Borji, Ali
    Itti, Laurent
    [J]. JOURNAL OF VISION, 2014, 14 (03):
  • [10] State-of-the-Art in Visual Attention Modeling
    Borji, Ali
    Itti, Laurent
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2013, 35 (01) : 185 - 207