Unsupervised domain adaptation with self-training for weed segmentation

被引:0
|
作者
Huang, Yingchao [1 ]
Hussein, Amina E. [2 ]
Wang, Xin [1 ]
Bais, Abdul [3 ]
Yao, Shanshan [4 ]
Wilder, Tanis [1 ]
机构
[1] Saskatchewan Polytech, Fac Digital Innovat Arts & Sci, Regina, SK S4S 5X1, Canada
[2] Univ Alberta, Dept Elect & Comp Engn, Edmonton, AB T6G 2H5, Canada
[3] Univ Regina, Fac Engn & Appl Sci, Elect Syst Engn, Regina, SK S4S 0A2, Canada
[4] Univ Alberta, Civil & Environm Engn, Edmonton, AB T6G 2H5, Canada
来源
基金
加拿大自然科学与工程研究理事会;
关键词
Weed identification; Semantic segmentation; Domain adaptation; Self-training; Convolutional neural networks; Dynamic weights; SEMANTIC SEGMENTATION; NEURAL-NETWORKS; CLASSIFICATION; MANAGEMENT;
D O I
10.1016/j.iswa.2024.200468
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Accurate crop and weed segmentation in varied field conditions is crucial for advancing automated weed management but remains challenging. Though promising, convolutional neural networks (CNNs) often experience performance drops when deployed in new field environments due to shifts between training and test data distributions. To address this limitation, we proposed a self-training framework using a teacher-student model that adapts CNNs for diverse agricultural contexts. Our method enhances generalization by co-training the student model on both the source domain and pseudo-labelled target domain generated by the teacher model, with teacher parameters updated via an exponential moving average of the student's model. The main contributions of this work areas follows: (1) we simplified the self-training procedure by using all target predictions, skipping the selection phase, and applying local dynamic weights (LDW) for target pixels during co-training; (2) we optimized iteration by monitoring covariance fluctuations to avoid pseudo-label overfitting and reduced the impact of false labels; (3) we addressed class imbalance with dynamic class weights (DCW) to give more importance to minority classes; and (4) we formulated a loss function integrating both LDW and DCW into the soft intersection over union (softIoU), enhancing weed segmentation effectiveness. We evaluated our framework with the ROSE challenge dataset across eight adaptations involving varied plants, robots, and growth stages, achieving up to a 0.17 mean IoU improvement over popular methods like CycleGAN. Our approach demonstrated consistent performance across diverse agricultural environments, supporting its use in real-field inference.
引用
收藏
页数:12
相关论文
共 50 条
  • [21] Unsupervised Domain Adaptation for 3D Object Detection via Self-Training
    Luo, Di
    PATTERN RECOGNITION AND COMPUTER VISION, PRCV 2023, PT II, 2024, 14426 : 307 - 318
  • [22] Geometry-Aware Self-Training for Unsupervised Domain Adaptation on Object Point Clouds
    Zou, Longkun
    Tang, Hui
    Chen, Ke
    Jia, Kui
    2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, : 6383 - 6392
  • [23] Understanding Self-Training for Gradual Domain Adaptation
    Kumar, Ananya
    Ma, Tengyu
    Liang, Percy
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 119, 2020, 119
  • [24] Back-Training excels Self-Training at Unsupervised Domain Adaptation of Question Generation and Passage Retrieval
    Kulshreshtha, Devang
    Belfer, Robert
    Serban, Iulian Vlad
    Reddy, Siva
    2021 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING (EMNLP 2021), 2021, : 7064 - 7078
  • [25] Single slice thigh CT muscle group segmentation with domain adaptation and self-training
    Yang, Qi
    Yu, Xin
    Lee, Ho Hin
    Cai, Leon Y.
    Xu, Kaiwen
    Bao, Shunxing
    Huo, Yuankai
    Moore, Ann Zenobia
    Makrogiannis, Sokratis
    Ferrucci, Luigi
    Landman, Bennett A.
    JOURNAL OF MEDICAL IMAGING, 2023, 10 (04)
  • [26] Domain Adaptation for Medical Image Segmentation Using Transformation-Invariant Self-training
    Ghamsarian, Negin
    Tejero, Javier Gamazo
    Marquez-Neila, Pablo
    Wolf, Sebastian
    Zinkernagel, Martin
    Schoeffmann, Klaus
    Sznitman, Raphael
    MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION, MICCAI 2023, PT I, 2023, 14220 : 331 - 341
  • [27] Unsupervised Domain Adaptation with Adversarial Self-Training for Crop Classification Using Remote Sensing Images
    Kwak, Geun-Ho
    Park, No-Wook
    REMOTE SENSING, 2022, 14 (18)
  • [28] Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-training
    Li, Zhenyu
    Chen, Zehui
    Li, Ang
    Fang, Liangji
    Jiang, Qinhong
    Liu, Xianming
    Jiang, Junjun
    COMPUTER VISION, ECCV 2022, PT IX, 2022, 13669 : 245 - 262
  • [29] Adversarial Self-Training with Domain Mask for Semantic Segmentation
    Hsin, Hsien-Kai
    Chiu, Hsiao-Chien
    Lin, Chun-Chen
    Chen, Chih-Wei
    Tsung, Pei-Kuei
    2019 IEEE INTELLIGENT TRANSPORTATION SYSTEMS CONFERENCE (ITSC), 2019, : 3689 - 3695
  • [30] Adversarial Domain Adaptation Enhanced via Self-training
    Altinel, Fazil
    Akkaya, Ibrahim Batuhan
    29TH IEEE CONFERENCE ON SIGNAL PROCESSING AND COMMUNICATIONS APPLICATIONS (SIU 2021), 2021,