| Title | A comparative evaluation of convolutional neural networks, training image sizes, and deep learning optimizers for weed detection in alfalfa |
| Authors | Yang, Jie Bagavathiannan, Muthukumar Wang, Yundi Chen, Yong Yu, Jialin |
| Affiliation | Nanjing Forestry Univ, Coll Mech & Elect Engn, Nanjing 210037, Peoples R China Peking Univ, Inst Adv Agr Sci, Weifang, Shandong, Peoples R China Texas A&M Univ, Dept Soil & Crop Sci, College Stn, TX 77843 USA Stevens Inst Technol, Dept Comp Sci, Hoboken, NJ 07030 USA Peking Univ, Inst Adv Agr Sci, Shandong Lab Adv Agr Sci Weifang, Weifang 261325, Shandong, Peoples R China |
| Keywords | SUPPORT VECTOR MACHINE CLASSIFICATION RECOGNITION CROP |
| Issue Date | Jun-2022 |
| Publisher | WEED TECHNOLOGY |
| Abstract | In this research, the deep-learning optimizers Adagrad, AdaDelta, Adaptive Moment Estimation (Adam), and Stochastic Gradient Descent (SGD) were applied to the deep convolutional neural networks AlexNet, GoogLeNet, VGGNet, and ResNet that were trained to recognize weeds among alfalfa using photographic images taken at 200x200, 400x400, 600x600, and 800x800 pixels. An increase in the image sizes reduced the classification accuracy of all neural networks. The neural networks that were trained with images of 200x200 pixels resulted in better classification accuracy than the other image sizes investigated here. The optimizers AlexNet and GoogLeNet trained with AdaDelta and SGD outperformed the Adagrad and Adam optimizers; VGGNet trained with AdaDelta outperformed Adagrad, Adam, and SGD; and ResNet trained with AdaDelta and Adagrad outperformed the Adam and SGD optimizers. When the neural networks were trained with the best-performing input image size (200x200 pixels) and the best-performing deep learning optimizer, VGGNet was the most effective neural network, with high precision and recall values (>= 0.99) when validation and testing datasets were used. Alternatively, ResNet was the least effective neural network in its ability to classify images containing weeds. However, there was no difference among the different neural networks in their ability to differentiate between broadleaf and grass weeds. The neural networks discussed herein may be used for scouting weed infestations in alfalfa and further integrated into the machine vision subsystem of smart sprayers for site-specific weed control. |
| URI | http://hdl.handle.net/20.500.11897/652679 |
| ISSN | 0890-037X |
| DOI | 10.1017/wet.2022.46 |
| Indexed | SCI(E) |
| Appears in Collections: | 待认领 |
