Attention Masking for Improved Near Out-of-Distribution Image Detection

被引:3
|
作者
Sim, Minho [1 ]
Lee, Jongwhoa [1 ]
Choi, Ho-Jin [1 ]
机构
[1] Korea Adv Inst Sci & Technol, Sch Comp, Daejeon, South Korea
关键词
out-of-distribution detection; image classification; vision transformer; quantifying attention map;
D O I
10.1109/BigComp57234.2023.00040
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
Detecting near out-of-distribution (OOD) data is important when deploying a deep learning model. The goal of near-OOD detection is to distinguish the OOD samples when distributions of inliers and outliers are similar. However, an input sample containing unexpected information may degrade the OOD detection performance in downstream tasks. To address this problem, we propose an algorithm called attention masking, which masks the less-attended parts of the given input to precisely calculate its Mahalanobis distance from the training distribution. In our experiments, we use a largescale pre-trained model to measure the performance of our approach on the vision OOD benchmark tasks. For instance, on CIFAR-100 vs. CIFAR-10 detection, we improve the AUROC of Mahalanobis distance-based OOD detection from 91.23% to 92.89% using ViT-Base model. In addition, we measured the performance of the challenging zero-shot OOD detection, which only utilizes the pre-trained weights without fine-tuning on CIFAR-100 or CIFAR-10, and achieved an average AUROC improvement of 7% on CIFAR-100 vs. CIFAR-10 detection.
引用
收藏
页码:195 / 202
页数:8
相关论文
共 50 条
  • [1] Out-of-Distribution Detection with Semantic Mismatch Under Masking
    Yang, Yijun
    Gao, Ruiyuan
    Xu, Qiang
    COMPUTER VISION, ECCV 2022, PT XXIV, 2022, 13684 : 373 - 390
  • [2] Leveraging Visual Attention for out-of-distribution Detection
    Cultrera, Luca
    Seidenari, Lorenzo
    Del Bimbo, Alberto
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS, ICCVW, 2023, : 4449 - 4458
  • [3] Research on Image Out-of-Distribution Detection: A Review
    Guo L.
    Li G.
    Gong K.
    Xue Z.
    Moshi Shibie yu Rengong Zhineng/Pattern Recognition and Artificial Intelligence, 2023, 36 (07): : 613 - 633
  • [4] Out-of-Distribution Detection with Uncertainty Enhanced Attention Maps
    Gao, Yue
    Su, Qinliang
    2021 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2021,
  • [5] Out-of-distribution Detection Learning with Unreliable Out-of-distribution Sources
    Zheng, Haotian
    Wang, Qizhou
    Fang, Zhen
    Xia, Xiaobo
    Liu, Feng
    Liu, Tongliang
    Han, Bo
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,
  • [6] Boosting Out-of-Distribution Image Detection With Epistemic Uncertainty
    Oh, Dokwan
    Ji, Daehyun
    Kwon, Ohmin
    Hyun, Yoonsuk
    IEEE ACCESS, 2022, 10 : 109289 - 109298
  • [7] Benchmarking Image Classifiers for Physical Out-of-Distribution Examples Detection
    Ojaswee
    Agarwal, Akshay
    Ratha, Nalini
    2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS, ICCVW, 2023, : 4429 - 4437
  • [8] An Efficient Data Augmentation Network for Out-of-Distribution Image Detection
    Lin, Cheng-Hung
    Lin, Cheng-Shian
    Chou, Po-Yung
    Hsu, Chen-Chien
    IEEE ACCESS, 2021, 9 : 35313 - 35323
  • [9] On the Learnability of Out-of-distribution Detection
    Fang, Zhen
    Li, Yixuan
    Liu, Feng
    Han, Bo
    Lu, Jie
    Journal of Machine Learning Research, 2024, 25
  • [10] Entropic Out-of-Distribution Detection
    Macedo, David
    Ren, Tsang Ing
    Zanchettin, Cleber
    Oliveira, Adriano L., I
    Ludermir, Teresa
    2021 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN), 2021,