Jun Peng
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jun Peng has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
Gui-Shu Wang
Department of Chemistry , Jiangxi Agricultural University , Nanchang 330031 , People’s Republic of China
1 shared publication- Haoyuan Yang
School of Computer and Information Engineering, Jiangxi Agricultural University, Nanchang 330045, China
1 shared publication Jia Gu
Department of Chemistry , Jiangxi Agricultural University , Nanchang 330045 , People’s Republic of China
1 shared publication- Jianjun Tang
College of Computer and Information Engineering, Jiangxi Agricultural University, Nanchang 330045, China
1 shared publication - Le Yang
School of Computer and Information Engineering, Jiangxi Agricultural University, Nanchang, China
1 shared publication - Liting Liao
School of Computer and Information Engineering, Jiangxi Agricultural University, Nanchang 330045, China
1 shared publication
Latest Publications
- Open Access
Remote Sensing Image Scene Classification with SE-EfficientNetV2-S: An Empirical Study of Channel Attention and Semi-Supervised Pseudo-Labeling
With the rapid development of remote sensing technology, high-resolution satellite imagery has been increasingly applied to land resource monitoring, urban planning, and environmental assessment. Automatically assigning semantic labels to remote sensing image patches remains a fundamental challenge due to pronounced intra-class variation and high inter-class visual similarity. To address the trade-off between model capacity and limited labeled data, this paper proposes a remote sensing image scene classification framework based on an improved EfficientNetV2-S architecture. The proposed model integrates a Squeeze-and-Excitation (SE) channel attention module between the final 1 × 1 expansion convolution and the Global Average Pooling layer, where it functions as a late-stage channel gating mechanism that adaptively recalibrates channel-wise responses, though its accuracy benefit is seed-sensitive rather than consistently reproducible at the current dataset scale. A two-stage optimization strategy was evaluated, comprising a fully unfrozen supervised baseline followed by a pseudo-label semi-supervised fine-tuning stage utilizing a strict confidence threshold (). Evaluated on a 10-class subset of the public NWPU-RESISC45 benchmark, the purely supervised SE-EfficientNetV2-S delivers 98.71% independent test accuracy, matching or exceeding the much larger ResNet50 (98.50%, 24.1 M parameters) despite using only 20.4 M parameters. Multi-seed variance analysis further reveals that semi-supervised fine-tuning yields a small test-set improvement for the No-SE configuration that is consistent in sign across all three seeds (+0.46 pp mean) but not statistically significant at this sample size, and an even smaller, likewise non-significant gain for the SE-augmented model (+0.08 pp), suggesting that channel gating moderates pseudo-label effectiveness in small-data regimes.Open AccessRemote Sensing Image Scene Classification with SE-EfficientNetV2-S: An Empirical Study of Channel Attention and Semi-Supervised Pseudo-Labeling
With the rapid development of remote sensing technology, high-resolution satellite imagery has been increasingly applied to land resource monitoring, urban planning, and environmental assessment. Automatically assigning semantic labels to remote sensing image patches remains a fundamental challenge due to pronounced intra-class variation and high inter-class visual similarity. To address the trade-off between model capacity and limited labeled data, this paper proposes a remote sensing image scene classification framework based on an improved EfficientNetV2-S architecture. The proposed model integrates a Squeeze-and-Excitation (SE) channel attention module between the final 1 × 1 expansion convolution and the Global Average Pooling layer, where it functions as a late-stage channel gating mechanism that adaptively recalibrates channel-wise responses, though its accuracy benefit is seed-sensitive rather than consistently reproducible at the current dataset scale. A two-stage optimization strategy was evaluated, comprising a fully unfrozen supervised baseline followed by a pseudo-label semi-supervised fine-tuning stage utilizing a strict confidence threshold (). Evaluated on a 10-class subset of the public NWPU-RESISC45 benchmark, the purely supervised SE-EfficientNetV2-S delivers 98.71% independent test accuracy, matching or exceeding the much larger ResNet50 (98.50%, 24.1 M parameters) despite using only 20.4 M parameters. Multi-seed variance analysis further reveals that semi-supervised fine-tuning yields a small test-set improvement for the No-SE configuration that is consistent in sign across all three seeds (+0.46 pp mean) but not statistically significant at this sample size, and an even smaller, likewise non-significant gain for the SE-augmented model (+0.08 pp), suggesting that channel gating moderates pseudo-label effectiveness in small-data regimes. - Open Access
- 5 November 2025
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 37 (9) , 1-25
Three-dimensional morphological reconstruction of potato leaf from a single image
- 5 November 2025
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 37 (9) , 1-25
Open AccessThree-dimensional morphological reconstruction of potato leaf from a single image
- Open Access
- 29 June 2022
- Published by Springer Nature in BMC Bioinformatics
- Vol. 23 (1) , 1-18
Hyb4mC: a hybrid DNA2vec-based model for DNA N4-methylcytosine sites prediction
Background: DNA N4-methylcytosine is part of the restrictive modification system, which works by regulating some biological processes, for example, the initiation of DNA replication, mismatch repair and inactivation of transposon. However, using experimental methods to detect 4mC sites is time-consuming and expensive. Besides, considering the huge differences in the number of 4mC samples among different species, it is challenging to achieve a robust multi-species 4mC site prediction performance. Hence, it is of great significance to develop effective computational tools to identify 4mC sites. Results: This work proposes a flexible deep learning-based framework to predict 4mC sites, called Hyb4mC. Hyb4mC adopts the DNA2vec method for sequence embedding, which captures more efficient and comprehensive information compared with the sequence-based feature method. Then, two different subnets are used for further analysis: Hyb_Caps and Hyb_Conv. Hyb_Caps is composed of a capsule neural network and can generalize from fewer samples. Hyb_Conv combines the attention mechanism with a text convolutional neural network for further feature learning. Conclusions: Extensive benchmark tests have shown that Hyb4mC can significantly enhance the performance of predicting 4mC sites compared with the recently proposed methods.
- 29 June 2022
- Published by Springer Nature in BMC Bioinformatics
- Vol. 23 (1) , 1-18
Open AccessHyb4mC: a hybrid DNA2vec-based model for DNA N4-methylcytosine sites prediction
Background: DNA N4-methylcytosine is part of the restrictive modification system, which works by regulating some biological processes, for example, the initiation of DNA replication, mismatch repair and inactivation of transposon. However, using experimental methods to detect 4mC sites is time-consuming and expensive. Besides, considering the huge differences in the number of 4mC samples among different species, it is challenging to achieve a robust multi-species 4mC site prediction performance. Hence, it is of great significance to develop effective computational tools to identify 4mC sites. Results: This work proposes a flexible deep learning-based framework to predict 4mC sites, called Hyb4mC. Hyb4mC adopts the DNA2vec method for sequence embedding, which captures more efficient and comprehensive information compared with the sequence-based feature method. Then, two different subnets are used for further analysis: Hyb_Caps and Hyb_Conv. Hyb_Caps is composed of a capsule neural network and can generalize from fewer samples. Hyb_Conv combines the attention mechanism with a text convolutional neural network for further feature learning. Conclusions: Extensive benchmark tests have shown that Hyb4mC can significantly enhance the performance of predicting 4mC sites compared with the recently proposed methods.
- Open Access
- 31 August 2021
- Published by Walter de Gruyter GmbH in Zeitschrift für Kristallographie - New Crystal Structures
- Vol. 236 (6) , 1215-1217
Crystal structure of methyl (E)-3-(4-acetoxyphenyl)acrylate, C12H12O4
C 12 H 12 O 4 , monoclinic, P 2 1 / c (no. 14), a = 16.071(6) Å, b = 9.424(4) Å, c = 7.640(3) Å, β = 101.861(5)°, V = 1132.4(8) Å 3 , Z = 4, R gt ( F ) = 0.0366, wR ref ( F 2 ) = 0.1043, T = 296(2) K.
- 31 August 2021
- Published by Walter de Gruyter GmbH in Zeitschrift für Kristallographie - New Crystal Structures
- Vol. 236 (6) , 1215-1217
Open AccessCrystal structure of methyl (E)-3-(4-acetoxyphenyl)acrylate, C12H12O4
C 12 H 12 O 4 , monoclinic, P 2 1 / c (no. 14), a = 16.071(6) Å, b = 9.424(4) Å, c = 7.640(3) Å, β = 101.861(5)°, V = 1132.4(8) Å 3 , Z = 4, R gt ( F ) = 0.0366, wR ref ( F 2 ) = 0.1043, T = 296(2) K.