Yan Xu
Yan Xu has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Qian Du
Mississippi State University
2 shared publications - Chiranjibi Shah
Northern Gulf Institute, Mississippi State University, Starkville, MS 39759, USA
1 shared publication - Guoming Li
Department of Poultry Science, the University of Georgia, Athens, GA 30602, USA
1 shared publication - Yanbo Huang
USDA Agricultural Research Service, Mississippi State, Mississippi, United States
1 shared publication - Yang Zhao
Department of Animal Science, University of Tennessee, Knoxville, TN 37996, USA
1 shared publication
Latest Publications
- Open Access
- 3 February 2022
- Published by MDPI AG in Remote Sensing
- Vol. 14 (3) , 716
Enhanced TabNet: Attentive Interpretable Tabular Learning for Hyperspectral Image Classification
Tree-based methods and deep neural networks (DNNs) have drawn much attention in the classification of images. Interpretable canonical deep tabular data learning architecture (TabNet) that combines the concept of tree-based techniques and DNNs can be used for hyperspectral image classification. Sequential attention is used in such architecture for choosing appropriate salient features at each decision step, which enables interpretability and efficient learning to increase learning capacity. In this paper, TabNet with spatial attention (TabNets) is proposed to include spatial information, in which a 2D convolution neural network (CNN) is incorporated inside an attentive transformer for spatial soft feature selection. In addition, spatial information is exploited by feature extraction in a pre-processing stage, where an adaptive texture smoothing method is used to construct a structure profile (SP), and the extracted SP is fed into TabNet (sTabNet) to further enhance performance. Moreover, the performance of TabNet-class approaches can be improved by introducing unsupervised pretraining. Overall accuracy for the unsupervised pretrained version of the proposed TabNets, i.e., uTabNets, can be improved from 11.29% to 12.61%, 3.6% to 7.67%, and 5.97% to 8.01% in comparison to other classification techniques, at the cost of increases in computational complexity by factors of 1.96 to 2.52, 2.03 to 3.45, and 2.67 to 5.52, respectively. Experimental results obtained on different hyperspectral datasets demonstrated the superiority of the proposed approaches in comparison with other state-of-the-art techniques including DNNs and decision tree variants.
- 3 February 2022
- Published by MDPI AG in Remote Sensing
- Vol. 14 (3) , 716
Open AccessEnhanced TabNet: Attentive Interpretable Tabular Learning for Hyperspectral Image Classification
Tree-based methods and deep neural networks (DNNs) have drawn much attention in the classification of images. Interpretable canonical deep tabular data learning architecture (TabNet) that combines the concept of tree-based techniques and DNNs can be used for hyperspectral image classification. Sequential attention is used in such architecture for choosing appropriate salient features at each decision step, which enables interpretability and efficient learning to increase learning capacity. In this paper, TabNet with spatial attention (TabNets) is proposed to include spatial information, in which a 2D convolution neural network (CNN) is incorporated inside an attentive transformer for spatial soft feature selection. In addition, spatial information is exploited by feature extraction in a pre-processing stage, where an adaptive texture smoothing method is used to construct a structure profile (SP), and the extracted SP is fed into TabNet (sTabNet) to further enhance performance. Moreover, the performance of TabNet-class approaches can be improved by introducing unsupervised pretraining. Overall accuracy for the unsupervised pretrained version of the proposed TabNets, i.e., uTabNets, can be improved from 11.29% to 12.61%, 3.6% to 7.67%, and 5.97% to 8.01% in comparison to other classification techniques, at the cost of increases in computational complexity by factors of 1.96 to 2.52, 2.03 to 3.45, and 2.67 to 5.52, respectively. Experimental results obtained on different hyperspectral datasets demonstrated the superiority of the proposed approaches in comparison with other state-of-the-art techniques including DNNs and decision tree variants.
- Open Access
Evaluating Convolutional Neural Networks for Cage-Free Floor Egg Detection
The manual collection of eggs laid on the floor (or ‘floor eggs’) in cage-free (CF) laying hen housing is strenuous and time-consuming. Using robots for automatic floor egg collection offers a novel solution to reduce labor yet relies on robust egg detection systems. This study sought to develop vision-based floor-egg detectors using three Convolutional Neural Networks (CNNs), i.e., single shot detector (SSD), faster region-based CNN (faster R-CNN), and region-based fully convolutional network (R-FCN), and evaluate their performance on floor egg detection under simulated CF environments. The results show that the SSD detector had the highest precision (99.9 ± 0.1%) and fastest processing speed (125.1 ± 2.7 ms·image−1) but the lowest recall (72.1 ± 7.2%) and accuracy (72.0 ± 7.2%) among the three floor-egg detectors. The R-FCN detector had the slowest processing speed (243.2 ± 1.0 ms·image−1) and the lowest precision (93.3 ± 2.4%). The faster R-CNN detector had the best performance in floor egg detection with the highest recall (98.4 ± 0.4%) and accuracy (98.1 ± 0.3%), and a medium prevision (99.7 ± 0.2%) and image processing speed (201.5 ± 2.3 ms·image−1); thus, the faster R-CNN detector was selected as the optimal model. The faster R-CNN detector performed almost perfectly for floor egg detection under a wide range of simulated CF environments and system settings, except for brown egg detection at 1 lux light intensity. When tested under random settings, the faster R-CNN detector had 91.9–94.7% precision, 99.8–100.0% recall, and 91.9–94.5% accuracy for floor egg detection. It is concluded that a properly-trained CNN floor-egg detector may accurately detect floor eggs under CF housing environments and has the potential to serve as a crucial vision-based component for robotic floor egg collection systems.
Open AccessEvaluating Convolutional Neural Networks for Cage-Free Floor Egg Detection
The manual collection of eggs laid on the floor (or ‘floor eggs’) in cage-free (CF) laying hen housing is strenuous and time-consuming. Using robots for automatic floor egg collection offers a novel solution to reduce labor yet relies on robust egg detection systems. This study sought to develop vision-based floor-egg detectors using three Convolutional Neural Networks (CNNs), i.e., single shot detector (SSD), faster region-based CNN (faster R-CNN), and region-based fully convolutional network (R-FCN), and evaluate their performance on floor egg detection under simulated CF environments. The results show that the SSD detector had the highest precision (99.9 ± 0.1%) and fastest processing speed (125.1 ± 2.7 ms·image−1) but the lowest recall (72.1 ± 7.2%) and accuracy (72.0 ± 7.2%) among the three floor-egg detectors. The R-FCN detector had the slowest processing speed (243.2 ± 1.0 ms·image−1) and the lowest precision (93.3 ± 2.4%). The faster R-CNN detector had the best performance in floor egg detection with the highest recall (98.4 ± 0.4%) and accuracy (98.1 ± 0.3%), and a medium prevision (99.7 ± 0.2%) and image processing speed (201.5 ± 2.3 ms·image−1); thus, the faster R-CNN detector was selected as the optimal model. The faster R-CNN detector performed almost perfectly for floor egg detection under a wide range of simulated CF environments and system settings, except for brown egg detection at 1 lux light intensity. When tested under random settings, the faster R-CNN detector had 91.9–94.7% precision, 99.8–100.0% recall, and 91.9–94.5% accuracy for floor egg detection. It is concluded that a properly-trained CNN floor-egg detector may accurately detect floor eggs under CF housing environments and has the potential to serve as a crucial vision-based component for robotic floor egg collection systems.