Class-Incremental Learning for Action Recognition in Videos

被引:2
|
作者
Park, Jaeyoo [1 ]
Kang, Minsoo [1 ]
Han, Bohyung [1 ]
机构
[1] Seoul Natl Univ, ECE & ASRI, Seoul, South Korea
基金
新加坡国家研究基金会;
关键词
D O I
10.1109/ICCV48922.2021.01344
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We tackle catastrophic forgetting problem in the context of class-incremental learning for video recognition, which has not been explored actively despite the popularity of continual learning. Our framework addresses this challenging task by introducing time-channel importance maps and exploiting the importance maps for learning the representations of incoming examples via knowledge distillation. We also incorporate a regularization scheme in our objective function, which encourages individual features obtained from different time steps in a video to be uncorrelated and eventually improves accuracy by alleviating catastrophic forgetting. We evaluate the proposed approach on brand-new splits of class-incremental action recognition benchmarks constructed upon the UCF101, HMDB51, and Something-Something V2 datasets, and demonstrate the effectiveness of our algorithm in comparison to the existing continual learning methods that are originally designed for image data.
引用
收藏
页码:13678 / 13687
页数:10
相关论文
共 50 条
  • [1] Class-Incremental Exemplar Compression for Class-Incremental Learning
    Luo, Zilin
    Liu, Yaoyao
    Schiele, Bernt
    Sun, Qianru
    [J]. 2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2023, : 11371 - 11380
  • [2] Class-Incremental Learning on Video-Based Action Recognition by Distillation of Various Knowledge
    Maraghi, Vali Ollah
    Faez, Karim
    [J]. COMPUTATIONAL INTELLIGENCE AND NEUROSCIENCE, 2022, 2022
  • [3] Class-Incremental Learning for Recognition of Complex-Valued Signals
    Fan, Zhaoyu
    Tu, Ya
    Lin, Yun
    Shi, Qingjiang
    [J]. IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING, 2024, 10 (02) : 417 - 428
  • [4] CLASS-INCREMENTAL LEARNING WITH REPETITION
    Hemati, Hamed
    Cossu, Andrea
    Carta, Antonio
    Hurtado, Julio
    Pellegrini, Lorenzo
    Bacciu, Davide
    Lomonaco, Vincenzo
    Borth, Damian
    [J]. CONFERENCE ON LIFELONG LEARNING AGENTS, VOL 232, 2023, 232 : 437 - 455
  • [5] Federated Class-Incremental Learning
    Dong, Jiahua
    Wang, Lixu
    Fang, Zhen
    Sun, Gan
    Xu, Shichao
    Wang, Xiao
    Zhu, Qi
    [J]. 2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 10154 - 10163
  • [6] Few-Shot Class-Incremental Learning for Named Entity Recognition
    Wang, Rui
    Yu, Tong
    Zhao, Handong
    Kim, Sungchul
    Mitra, Subrata
    Zhang, Ruiyi
    Henao, Ricardo
    [J]. PROCEEDINGS OF THE 60TH ANNUAL MEETING OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS (ACL 2022), VOL 1: (LONG PAPERS), 2022, : 571 - 582
  • [7] Class-Incremental Gesture Recognition Learning with Out-of-Distribution Detection
    Li, Mingxue
    Cong, Yang
    Liu, Yuyang
    Sun, Gan
    [J]. 2022 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2022, : 1503 - 1508
  • [8] Is Class-Incremental Enough for Continual Learning?
    Cossu, Andrea
    Graffieti, Gabriele
    Pellegrini, Lorenzo
    Maltoni, Davide
    Bacciu, Davide
    Carta, Antonio
    Lomonaco, Vincenzo
    [J]. FRONTIERS IN ARTIFICIAL INTELLIGENCE, 2022, 5
  • [9] Deep Learning for Class-Incremental Learning: A Survey
    Zhou, Da-Wei
    Wang, Fu-Yun
    Ye, Han-Jia
    Zhan, De-Chuan
    [J]. Jisuanji Xuebao/Chinese Journal of Computers, 2023, 46 (08): : 1577 - 1605
  • [10] AdvisIL - A Class-Incremental Learning Advisor
    Feillet, Eva
    Petit, Gregoire
    Popescu, Adrian
    Reyboz, Marina
    Hudelot, Celine
    [J]. 2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), 2023, : 2399 - 2408