Learning Canonical 3D Object Representation for Fine-Grained Recognition

被引:11
|
作者
Joung, Sunghun [1 ]
Kim, Seungryong [2 ]
Kim, Minsu [1 ]
Kim, Ig-Jae [3 ]
Sohn, Kwanghoon [1 ]
机构
[1] Yonsei Univ, Seoul, South Korea
[2] Korea Univ, Seoul, South Korea
[3] Korea Inst Sci & Technol KIST, Seoul, South Korea
来源
2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021) | 2021年
基金
新加坡国家研究基金会;
关键词
D O I
10.1109/ICCV48922.2021.00107
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We propose a novel framework for fine-grained object recognition that learns to recover object variation in 3D space from a single image, trained on an image collection without using any ground-truth 3D annotation. We accomplish this by representing an object as a composition of 3D shape and its appearance, while eliminating the effect of camera viewpoint, in a canonical configuration. Unlike conventional methods modeling spatial variation in 2D images only, our method is capable of reconfiguring the appearance feature in a canonical 3D space, thus enabling the subsequent object classifier to be invariant under 3D geometric variation. Our representation also allows us to go beyond existing methods, by incorporating 3D shape variation as an additional cue for object recognition. To learn the model without ground-truth 3D annotation, we deploy a differentiable renderer in an analysis-by-synthesis framework. By incorporating 3D shape and appearance jointly in a deep representation, our method learns the discriminative representation of the object and achieves competitive performance on fine-grained image recognition and vehicle re-identification. We also demonstrate that the performance of 3D shape reconstruction is improved by learning fine-grained shape deformation in a boosting manner.
引用
收藏
页码:1015 / 1025
页数:11
相关论文
共 50 条
  • [1] LEARNING DEEP AND SPARSE FEATURE REPRESENTATION FOR FINE-GRAINED OBJECT RECOGNITION
    Srinivas, M.
    Lin, Yen-Yu
    Liao, Hong-Yuan Mark
    2017 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO (ICME), 2017, : 1458 - 1463
  • [2] 3D Object Representations for Fine-Grained Categorization
    Krause, Jonathan
    Stark, Michael
    Deng, Jia
    Li Fei-Fei
    2013 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCVW), 2013, : 554 - 561
  • [3] Towards Fine-Grained Recognition: Joint Learning for Object Detection and Fine-Grained Classification
    Wang, Qiaosong
    Rasmussen, Christopher
    ADVANCES IN VISUAL COMPUTING, ISVC 2019, PT II, 2019, 11845 : 332 - 344
  • [4] 3D Pose Estimation for Fine-Grained Object Categories
    Wang, Yaming
    Tan, Xiao
    Yang, Yi
    Liu, Xiao
    Ding, Errui
    Zhou, Feng
    Davis, Larry S.
    COMPUTER VISION - ECCV 2018 WORKSHOPS, PT I, 2019, 11129 : 619 - 632
  • [5] FINE-GRAINED AND LAYERED OBJECT RECOGNITION
    Wu, Yang
    Zheng, Nanning
    Liu, Yuanliu
    Yuan, Zejian
    INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE, 2012, 26 (02)
  • [6] A Refined 3D Pose Dataset for Fine-Grained Object Categories
    Wang, Yaming
    Tan, Xiao
    Yang, Yi
    Li, Ziyu
    Liu, Xiao
    Zhou, Feng
    Davis, Larry S.
    2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION WORKSHOPS (ICCVW), 2019, : 2797 - 2806
  • [7] LEVERAGING 2D AND 3D CUES FOR FINE-GRAINED OBJECT CLASSIFICATION
    Wang, Xiaolong
    Li, Robert
    Currey, Jon
    2016 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2016, : 1354 - 1358
  • [8] Fine-grained Object Recognition via Pose Alignment and Part based Representation
    Chen, Shuxian
    Liu, Jianming
    NINTH INTERNATIONAL CONFERENCE ON DIGITAL IMAGE PROCESSING (ICDIP 2017), 2017, 10420
  • [9] Fine-Grained Object Recognition with Gnostic Fields
    Kanan, Christopher
    2014 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), 2014, : 23 - 30
  • [10] JOINT LEARNING ON THE HIERARCHY REPRESENTATION FOR FINE-GRAINED HUMAN ACTION RECOGNITION
    Leong, Mei Chee
    Tan, Hui Li
    Zhang, Haosong
    Li, Liyuan
    Lin, Feng
    Lim, Joo Hwee
    2021 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2021, : 1059 - 1063