Fig 1.
Architecture of GAN [41].
Fig 2.
a) Generator. b) Discriminator.
Fig 3.
Overview of proposed methodology.
Fig 4.
Overview of UAV and experimental site.
a) DJI Spark (Multi-Rotor). b) Pea Field. c) Strawberry Field.
Fig 5.
Sample images for the crops (pea and strawberry) and weeds dataset.
a) Weeds. b) Crops. c) Crops.
Fig 6.
Crops and weeds classification for the developed system.
a) Weeds. b) Crops.
Table 1.
Number of the labeled and unlabeled Images under different labeled rates for two different croplands.
Table 2.
Average training time at different labeled rates.
Fig 7.
Cross validation accuracy (SSCA) using different labeled rates.
(a) Labeled Rate 20% b) Labeled rate 40% c) Labeled rate 60% d) Labeled rate 80%.
Table 3.
Classification accuracy of our proposed semi-supervised learning method under different labeled rates.
Table 4.
Supervised classification accuracy (SCA) of KNN, SVM, LeNet5, VGG-11VGG-16, ResNet18, and ResNet50 on different labeled rates.
Fig 8.
Comparison between SCA and SSCA at different labeled rates.
a) Pea. b) Strawberry.
Fig 9.
Comparison of developed system with individual systems.
a) Developed system vs KNN. b) Developed system vs SVM. c) Developed system vs LeNet5. d) Developed system vs VGG-11. e) Developed system vs VGG-16. f) Developed system vs ResNet18. g) Developed system vs ResNet50.
Fig 10.
Comparison of developed system with individual systems for strawberry.
a) Developed system vs KNN. b) Developed system vs SVM. c) Developed system vs LeNet5. d) Developed system vs VGG-11. e) Developed system vs VGG-16. f) Developed system vs ResNet18. g) Developed system vs ResNet50.