Article
Plant Sciences
Kui Hu, YongMin Liu, Jiawei Nie, Xinying Zheng, Wei Zhang, Yuan Liu, TianQiang Xie
Summary: The study proposes a multi-scale dual-branch structural rice pest identification model based on a generative adversarial network and improved ResNet. By using data preprocessing and enhancement methods, the dataset is expanded from 5,932 to 20,000, and the recognition accuracy of the model improves by 2.66% compared to the original ResNet model. The model demonstrates good generalization ability and robustness, providing a superior solution for crop pest and disease identification.
FRONTIERS IN PLANT SCIENCE
(2023)
Article
Engineering, Electrical & Electronic
Tanmay Anand, Soumendu Sinha, Murari Mandal, Vinay Chamola, Fei Richard Yu
Summary: Aerial inspection of agricultural regions provides crucial information to safeguard efficient farming, while monitoring farmland anomalies is essential for increasing agricultural technology efficiency and developing AI-assisted farming models. The proposal of the deep learning framework AgriSegNet contributes to automated detection of farmland anomalies and enhancing precision farming techniques.
IEEE SENSORS JOURNAL
(2021)
Article
Chemistry, Multidisciplinary
Muhammad Yaqub, Feng Jinchao, Shahzad Ahmed, Kaleem Arshid, Muhammad Atif Bilal, Muhammad Pervez Akhter, Muhammad Sultan Zia
Summary: Generative adversarial networks (GAN), powered by deep learning, are effective for image reconstruction using under-sampled MR data. This research explores the applications of deep learning-based GAN and transfer learning, achieving superior results in MRI reconstruction for brain and knee imaging.
APPLIED SCIENCES-BASEL
(2022)
Article
Agriculture, Multidisciplinary
Chengsong Hu, J. Alex Thomasson, Muthukumar V. Bagavathiannan
Summary: The study proposed a novel image synthesis and semi-supervised learning pipeline for training weed detection models without the need for manually labeled images, achieving performance levels close to that of supervised models. The results showed that color match, training-time color augmentation, and iterative semi-supervised learning significantly improved model performance.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2021)
Article
Plant Sciences
Satinder Bal Gupta, RajKumar Yadav, Fatemeh Bovand, Pankaj Kumar Tyagi
Summary: Castor is an important nonedible industrial crop that produces oil and is used in various products. Insect pest attacks can degrade the quality and quantity of castor oil. Traditional methods of identifying pests are time-consuming and require expertise. This study proposes a hybrid data augmentation approach to solve the lack of a suitable dataset for effective vision-based model training, using deep convolutional neural networks. The results show that this method improves overall performance compared to previous methods.
FRONTIERS IN PLANT SCIENCE
(2023)
Article
Chemistry, Analytical
Longzhe Quan, Bing Wu, Shouren Mao, Chunjie Yang, Hengda Li
Summary: The study presents a weed segmentation method based on BlendMask that effectively obtains phenotypic information of weeds under complex field conditions. By utilizing data enhancement and ResNet101 as the backbone network, the model performance was improved significantly. The deep learning method demonstrated great potential in accurately identifying weed leaf age and plant center, which is essential for variable spraying applications.
Article
Computer Science, Artificial Intelligence
Jinbao Wang, Guoyang Xie, Yawen Huang, Jiayi Lyu, Feng Zheng, Yefeng Zheng, Yaochu Jin
Summary: Utilizing multi-modal neuroimaging data is effective in studying human cognitive activities and pathologies, but obtaining full sets of centrally collected paired data is impractical. Federated learning is needed to integrate dispersed data from different institutions. The proposed FedMed-GAN algorithm bridges the gap between federated learning and medical GAN, mitigating mode collapse without sacrificing generator performance. It outperforms state-of-the-art methods in comprehensive evaluations.
Article
Mathematics
Subhajit Chatterjee, Debapriya Hazra, Yung-Cheol Byun, Yong-Woon Kim
Summary: Plastic bottle recycling is crucial for environmental protection. Using deep learning techniques for automatic classification can improve accuracy and reduce cost. The study proposes a GAN-based model for generating synthetic images and a modified lightweight-GAN model for enhancing image synthesis quality. The weighted average ensemble model based on pre-trained models achieves high classification accuracy.
Article
Computer Science, Information Systems
Shantam Shorewala, Armaan Ashfaque, R. Sidharth, Ujjwal Verma
Summary: This study proposes a deep learning-based semi-supervised approach to accurately estimate weed density and distribution across farmlands, facilitating selective treatment of weeds by autonomous robots.
Article
Computer Science, Artificial Intelligence
Haipeng Deng, Qiuxia Wu, Han Huang, Xiaowei Yang, Zhiyong Wang
Summary: The unsupervised image-to-image translation aims to learn a mapping that translates images from one domain to another. Current GAN models require expensive operations and suffer from high computational costs. To address this, we propose using involution, a lightweight operator, to enhance the GAN structure and introduce a novel loss term to evaluate perceptual similarity distance.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Aurele Tohokantche Aurele Tohokantche, Wenming Cao, Xudong Mao, Si Wu, Hau-San Wong, Qing Li
Summary: This paper proposes a parametric and robust AB loss function to improve the performance of generative adversarial networks (GAN) on different datasets and alleviate the issue of mode collapse. Experimental results demonstrate that this approach can enhance the quality of synthetic images.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Yuan Li, Jian Wang, Weibo Liang, Hui Xue, Zhenan He, Jiancheng Lv, Lin Zhang
Summary: This study proposes a novel craniofacial reconstruction method to identify human remains by synthesizing corresponding craniofacial images from 2D skull images. The method utilizes deep generative adversarial nets to automatically transform skull data to facial images. Experimental results demonstrate that the method can be used as an effective forensic tool.
PATTERN RECOGNITION
(2022)
Article
Agronomy
Tavseef Mairaj Shah, Durga Prasad Babu Nasika, Ralf Otterpohl
Summary: Farming systems play a crucial role in the world food production, which is directly linked to social, economic, and ecological systems. Weeds are identified as a major factor causing yield gaps in different regions worldwide. A plant and weed identifier tool based on artificial deep neural networks was developed to address the weed infestation issue in farming systems, achieving high accuracy in plant and weed prediction tasks.
Article
Environmental Sciences
Joseph K. Mhango, Edwin W. Harris, Richard Green, James M. Monaghan
Summary: This study utilized the Faster Region-based Convolutional Neural Network (FRCNN) framework to produce a plant detection model and estimate plant densities using UAV imagery, showing the accurate construction of two-dimensional maps of plant density with high correlation to important yield components. Despite the challenges of inaccurate predictions in images of merged canopies, the FRCNN model proved to be effective in predicting plant density and its relationship with potato yield attributes.
Article
Chemistry, Analytical
Furkat Safarov, Kuchkorov Temurbek, Djumanov Jamoljon, Ochilov Temur, Jean Chamberlain Chedjou, Akmalbek Bobomirzaevich Abdusalomov, Young-Im Cho
Summary: The study proposes a TL-ResUNet model for land cover classification and segmentation using satellite images, which combines the strengths of residual network, transfer learning, and UNet architecture. The model outperforms classic models on accuracy and performance, achieving good performance on the DeepGlobe dataset.
Article
Agricultural Engineering
Yifan Bai, Junzhen Yu, Shuqin Yang, Jifeng Ning
Summary: A real-time recognition algorithm (Improved YOLO) is proposed in this paper for accurately identifying small, similar-colored, and overlapping strawberry seedling flowers and fruits. The experimental results show that the algorithm achieves high precision, recall, and average precision, and meets the real-time detection requirements, providing effective support for the automated management of strawberry seedling flower and fruit thinning.
BIOSYSTEMS ENGINEERING
(2024)