Object Detection in Tomato Greenhouses: A Study on Model Generalization

被引:1
|
作者
Haggag, Sammar [1 ]
Veres, Matthew [1 ]
Tarry, Cole [1 ]
Moussa, Medhat [1 ]
机构
[1] Univ Guelph, Sch Engn, Guelph, ON N1G 2W1, Canada
来源
AGRICULTURE-BASEL | 2024年 / 14卷 / 02期
关键词
deep learning; generalization; object detection; greenhouse;
D O I
10.3390/agriculture14020173
中图分类号
S3 [农学(农艺学)];
学科分类号
0901 ;
摘要
Harvesting operations in agriculture are labour-intensive tasks. Automated solutions can help alleviate some of the pressure faced by rising costs and labour shortage. Yet, these solutions are often difficult and expensive to develop. To enable the use of harvesting robots, machine vision must be able to detect and localize target objects in a cluttered scene. In this work, we focus on a subset of harvesting operations, namely, tomato harvesting in greenhouses, and investigate the impact that variations in dataset size, data collection process and other environmental conditions may have on the generalization ability of a Mask-RCNN model in detecting two objects critical to the harvesting task: tomatoes and stems. Our results show that when detecting stems from a perpendicular perspective, models trained using data from the same perspective are similar to one that combines both perpendicular and angled data. We also show larger changes in detection performance across different dataset sizes when evaluating images collected from an angled camera perspective, and overall larger differences in performance when illumination is the primary source of variation in the data. These findings can be used to help practitioners prioritize data collection and evaluation efforts, and lead to larger-scale harvesting dataset construction efforts.
引用
收藏
页数:12
相关论文
共 50 条
  • [1] Domain Attention Model for Domain Generalization in Object Detection
    He, Weixiong
    Zheng, Huicheng
    Lai, Jianhuang
    PATTERN RECOGNITION AND COMPUTER VISION (PRCV 2018), PT IV, 2018, 11259 : 27 - 39
  • [2] Object-Aware Domain Generalization for Object Detection
    Lee, Wooju
    Hong, Dasol
    Lim, Hyungtae
    Myung, Hyun
    THIRTY-EIGHTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 38 NO 4, 2024, : 2947 - 2955
  • [3] A Multi-Modal Open Object Detection Model for Tomato Leaf Diseases with Strong Generalization Performance Using PDC-VLD
    Li, Jinyang
    Zhao, Fengting
    Zhao, Hongmin
    Zhou, Guoxiong
    Xu, Jiaxin
    Gao, Mingzhou
    Li, Xin
    Dai, Weisi
    Zhou, Honliang
    Hu, Yahui
    He, Mingfang
    PLANT PHENOMICS, 2024, 6
  • [4] Automated spider mite damage detection on tomato leaves in greenhouses
    Nieuwenhuizen, A. T.
    Kool, J.
    Suh, H. K.
    Hemming, J.
    XI INTERNATIONAL SYMPOSIUM ON PROTECTED CULTIVATION IN MILD WINTER CLIMATES AND I INTERNATIONAL SYMPOSIUM ON NETTINGS AND SCREENS IN HORTICULTURE, 2020, 1268 : 165 - 172
  • [5] Tomato Fruit Detection and Counting in Greenhouses Using Deep Learning
    Afonso, Manya
    Fonteijn, Hubert
    Fiorentin, Felipe Schadeck
    Lensink, Dick
    Mooij, Marcel
    Faber, Nanne
    Polder, Gerrit
    Wehrens, Ron
    FRONTIERS IN PLANT SCIENCE, 2020, 11
  • [6] RIPE TOMATO DETECTION FOR ROBOTIC VISION HARVESTING SYSTEMS IN GREENHOUSES
    Yin, H.
    Chai, Y.
    Yang, S. X.
    Mittal, G. S.
    TRANSACTIONS OF THE ASABE, 2011, 54 (04) : 1539 - 1546
  • [7] Tompousse, a model of yield prediction for tomato crops: Calibration study for unheated plastic greenhouses
    Abreu, P
    Meneses, JF
    Gary, C
    PROCEEDINGS OF THE XXV INTERNATIONAL HORTICULTURAL CONGRESS, PT 9: COMPUTERS AND AUTOMATION ELECTRONIC INFORMATION IN HORTICULTURE, 2000, (519): : 141 - 149
  • [8] TOMATO CULTURES IN GREENHOUSES
    BAETS, W
    VANDERBRUGGEN, M
    CEULEMANS, R
    IMPENS, I
    REVUE DE L AGRICULTURE, 1985, 38 (05): : 1127 - 1149
  • [9] Early detection and spread of Tomato Powdery Mildew (TPM) in commercial greenhouses
    Sokolidi, A.
    PHYTOPATHOLOGY, 2022, 112 (11) : 49 - 49
  • [10] Automatic Detection of the Main Vine and Branches of Tomato Plants Grown in Greenhouses
    El Mesalami, Ayman M.
    Ibrahim, Soad F.
    Moussa, Medhat
    2018 IEEE INTERNATIONAL CONFERENCE ON ELECTRO/INFORMATION TECHNOLOGY (EIT), 2018, : 753 - 758