Edge-Channel Aggregation Network and Two-Stage Fine Tuning Scheme for Handwritten Dongba Character Recognition

Xiali Li , Yang Xiao , Xiaojun Bi , Yanlong Luo

CAAI Transactions on Intelligence Technology ›› 2026, Vol. 11 ›› Issue (4) : 1254 -1273.

PDF (3054KB)
CAAI Transactions on Intelligence Technology ›› 2026, Vol. 11 ›› Issue (4) :1254 -1273. DOI: 10.1049/cit2.70161
ORIGINAL RESEARCH
research-article
Edge-Channel Aggregation Network and Two-Stage Fine Tuning Scheme for Handwritten Dongba Character Recognition
Author information +
History +
PDF (3054KB)

Abstract

Handwritten Dongba Character Recognition (HDCR) contains a large number of visually similar characters with subtle and fragile edge cues, posing severe challenges to feature learning. To address this issue, an Edge Channel Aggregation Network (EdgeCANet) model is proposed. The core of EdgeCANet is the Edge Channel Attention Aggregation (ECAA) block, which integrates the Edge-Guided Enhancement Module (EGEM) for edge-aware spatial refinement and the Channel Texture Recalibration Module (CTRM) for channel-wise texture recalibration. Besides, the two-stage misclassified sample weight redistribution fine-tuning scheme based on Exponential Moving Average (EMA) algorithm is proposed to stabilise parameter updates, and to emphasise difficult and highly similar classes. The recognition accuracy of EdgeCANet significantly outperforms state-of-the-art models across the constructed high-similarity datasets HS-DB and HS-S20 K, as well as the public DB1404, OBC306, Sketch-20 K, and ImageNet-Sketch datasets. The results validate that EdgeCANet effectively enhances recognition accuracy for highly similar handwritten characters and ancient minority-script characters, offering more promising applicability for related recognition tasks.

Keywords

channel attention mechanism / Dongba character recognition / edge attention mechanism / exponential moving average / fine tuning / image recognition

Cite this article

Download citation ▾
Xiali Li, Yang Xiao, Xiaojun Bi, Yanlong Luo. Edge-Channel Aggregation Network and Two-Stage Fine Tuning Scheme for Handwritten Dongba Character Recognition. CAAI Transactions on Intelligence Technology, 2026, 11 (4) : 1254-1273 DOI:10.1049/cit2.70161

登录浏览全文

4963

注册一个新账户 忘记密码

References

[1]

P. He, W. Zhang, H. Yang, and M. Zeng, Yunnan Minority Literature Picture Book Collection Genesis Chinese Naxi Hieroglyphics in English (Yunnan People’s Publishing House, 2021).

[2]

Y. Luo, Y. Sun, and X. Bi, “ Multiple Attentional Aggregation Network for Handwritten Dongba Character Recognition,” Expert Systems with Applications 213 (2023): 118865, https://doi.org/10.1016/j.eswa.2022.118865.

[3]

M. Eitz, J. Hays, and M. Alexa, “ How do Humans Sketch Objects?,” ACM Transactions on Graphics 31, no. 4 (2012): 1-10, https://doi.org/10.1145/2185520.2335395.

[4]

S. Huang, H. Wang, Y. Liu, et al., “ Obc306: A Large-Scale Oracle Bone Character Recognition Dataset,” in 2019 International Conference on Document Analysis and Recognition (ICDAR) (IEEE, 2019), 681-688.

[5]

S. Woo, S. Debnath, R. Hu, et al., “ Convnext v2: Co-Designing and Scaling Convnets With Masked Autoencoders,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2023), 16133-16142.

[6]

W. Wang, J. Dai, Z. Chen, et al., “ Internimage: Exploring Large-Scale Vision Foundation Models With Deformable Convolutions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2023), 14408-14419.

[7]

R. Hua and X. Xu, “ Intelligent Classification on Images of Dongba Ancient Books,” Journal of Engineering 2019, no. 23 (2019): 9039-9042, https://doi.org/10.1049/joe.2018.9177.

[8]

Y. Luo, X. Bi, L. Wu, et al., “ Recognition of Dongba Pictographs Based on Improved Residual Learning,” CAAI Transactions on Intelligent Systems 17, no. 1 (2021): 79-87, https://doi.org/10.11992/tis.202112009.

[9]

H. Luo, D. Xu, B. Yang, et al., “ Multi-Scale Feature Fusion Based Dongba Character Recognition,” in 2020 5th International Conference on Mechanical, Control and Computer Engineering (ICMCCE) (IEEE, 2020), 1571-1575.

[10]

X. Bi and Y. Luo, “ Incomplete Handwritten Dongba Character Image Recognition by Multiscale Feature Restoration,” Heritage Science 12, no. 1 (2024): 1-19, https://doi.org/10.1186/s40494-024-01329-8.

[11]

J. Li, X. Chi, Q. Wang, et al., “ A Comprehensive Survey of Oracle Character Recognition: Challenges, Datasets, Methodology, and Beyond,” Pattern Recognition 169 (2026): 111824, https://doi.org/10.1016/j.patcog.2025.111824.

[12]

M. A. Ko and S. Poruran, “ OCR-nets: Variants of Pre-Trained CNN for Urdu Handwritten Character Recognition via Transfer Learning,” Procedia Computer Science 171 (2020): 2294-2301, https://doi.org/10.1016/j.procs.2020.04.248.

[13]

M. A. Kilvisharam Oziuddeen, S. Poruran, and M. Y. Caffiyar, “ A Novel Deep Convolutional Neural Network Architecture Based on Transfer Learning for Handwritten Urdu Character Recognition,” Tehnički vjesnik 27, no. 4 (2020): 1160-1165, https://doi.org/10.17559/TV-20190319095323.

[14]

A. M. Kumar, J. I. R. Praveen, S. Venkatraman, et al., “ Improved Tomato Leaf Disease Classification Through Adaptive Ensemble Models With Exponential Moving Average Fusion and Enhanced Weighted Gradient Optimization,” Frontiers in Plant Science 15 (2024): 1382416, https://doi.org/10.3389/fpls.2024.1382416.

[15]

H. Wu, Y. Geng, D. Gai, et al., “ Dual-Student Adversarial Framework With Discriminator and Consistency-Driven Learning for Semi-Supervised Medical Image Segmentation,” IEEE Journal of Biomedical and Health Informatics 30, no. 2 (2026): 1492-1505, https://doi.org/10.1109/jbhi.2025.3597469.

[16]

He K., Fan H., Wu Y., et al., “ Momentum Contrast for Unsupervised Visual Representation Learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2020), 9729-9738.

[17]

A. Tarvainen, “ Mean Teachers are Better Role Models: Weight-Averaged Consistency Targets Improve Semi-Supervised Deep Learning Results,” Advances in Neural Information Processing Systems 30 (2017): 1196, https://dl.acm.org/doi/10.5555/3294771.3294885.

[18]

L. Wang, S. Chen, L. Jiang, et al., “ Parameter-Efficient Fine-Tuning in Large Language Models: A Survey of Methodologies,” Artificial Intelligence Review 58, no. 8 (2025): 227, https://doi.org/10.1007/s10462-025-11236-4.

[19]

T. ValizadehAslani, Y. Shi, J. Wang, et al., “ Two-Stage Fine-Tuning With Chatgpt Data Augmentation for Learning Class-Imbalanced Data,” Neurocomputing 592 (2024): 127801, https://doi.org/10.1016/j.neucom.2024.127801.

[20]

S. Bondugula, G. Qian, and A. Beach, “ Two-Stage Training for Improved Classification of Poorly Localized Object Images,” in European Conference on Computer Vision (Springer International Publishing, 2020), 245-260, https://link.springer.com/chapter/10.1007/978-3-030-68238-5_18.

[21]

M. Alsaidi, A. S. Altaher, M. T. Jan, et al., “ Covid-19 Classification Using Deep Learning Two-Stage Approach,” arXiv preprint, arXiv:2211.15817, November 28, 2022, https://doi.org/10.48550/arXiv.2211.15817.

[22]

T. Y. Lin, P. Goyal, R. Girshick, et al., “ Focal Loss for Dense Object Detection,” in Proceedings of the IEEE International Conference on Computer Vision (2017), 2980-2988.

[23]

K. Cao, C. Wei, A. Gaidon, et al., “ Learning Imbalanced Datasets With Label-Distribution-Aware Margin Loss,” Advances in Neural Information Processing Systems (2019): 32.

[24]

X. Yuan, X. Xu, X. Wang, et al., “ Osap-Loss: Efficient Optimization of Average Precision via Involving Samples After Positive Ones Towards Remote Sensing Image Retrieval,” CAAI Transactions on Intelligence Technology 8, no. 4 (2023): 1191-1212, https://doi.org/10.1049/cit2.12151.

[25]

C. Shorten and T. M. Khoshgoftaar, “ A Survey on Image Data Augmentation for Deep Learning,” Journal of Big Data 6, no. 1 (2019): 1-48, https://doi.org/10.1186/s40537-019-0197-0.

[26]

C. Luo, Y. Zhu, L. Jin, et al., “ Learn to Augment: Joint Data Augmentation and Network Optimization for Text Recognition,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2020), 13746-13755.

[27]

P. Alimisis, I. Mademlis, P. Radoglou-Grammatikis, P. Sarigiannidis, and G. T. Papadopoulos, “ Advances in Diffusion Models for Image Data Augmentation: A Review of Methods, Models, Evaluation Metrics and Future Research Directions,” Artificial Intelligence Review 58, no. 4 (2025): 112, https://doi.org/10.1007/s10462-025-11116-x.

[28]

X. Xiong, Y. Sun, X. Liu, et al., “ Distance Guided Generative Adversarial Network for Explainable Medical Image Classifications,” Computerized Medical Imaging and Graphics 118 (2024): 102444, https://doi.org/10.1016/j.compmedimag.2024.102444.

[29]

C. Zheng, G. Wu, and C. Li, “ Toward Understanding Generative Data Augmentation,” Advances in Neural Information Processing Systems 36 (2023): 54046-54060, https://doi.org/10.52202/075280-2352.

[30]

K. Islam, M. Z. Zaheer, A. Mahmood, et al., “ Diffusemix: Label-Preserving Data Augmentation With Diffusion Models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2024), 27621-27630.

[31]

T. Takikawa, D. Acuna, V. Jampani, et al., “ Gated-Scnn: Gated Shape Cnns for Semantic Segmentation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision (2019), 5229-5238.

[32]

N. T. Bui, D. H. Hoang, Q. T. Nguyen, et al., “ Meganet: Multi-Scale Edge-Guided Attention Network for Weak Boundary Polyp Segmentation,” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (2024), 7985-7994.

[33]

J. Hu, L. Shen, and G. Sun, “ Squeeze-And-Excitation Networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2018), 7132-7141.

[34]

Y. Huang, F. Liu, and A. G. S. M. C. Zhou, “ A Global-Local Scalable Multi-Task Contrastive Learning Framework,” in Computer Graphics International Conference (Springer Nature Switzerland, 2024), 297-308.

[35]

F. Liu, Z. Fu, and Y. Wang, “ Reward-Based Gradient Modulation for Multimodal Emotion Recognition With Lora,” IEEE Transactions on Computational Social Systems 12, no. 5 (2025): 1-11, https://doi.org/10.1109/tcss.2025.3566373.

[36]

A. S. Sambyal, U. Niyaz, S. Shrivastava, et al., “ Ls+: Informed Label Smoothing for Improving Calibration in Medical Image Classification,” in International Conference on Medical Image Computing and Computer-Assisted Intervention (Springer Nature Switzerland, 2024), 513-523.

[37]

B. Huang, H. Sharma, and Y. Mao, “ Enhancing Language Model Alignment: A Confidence-Based Approach to Label Smoothing,” in Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (2024), 21341-21352.

[38]

K. He, X. Zhang, S. Ren, et al., “ Deep Residual Learning for Image Recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2016), 770-778.

[39]

G. Huang, Z. Liu, L. Van Der Maaten, et al., “ Densely Connected Convolutional Networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017), 4700-4708.

[40]

S. Xie, R. Girshick, P. Dollár, et al., “ Aggregated Residual Transformations for Deep Neural Networks,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017), 1492-1500.

[41]

N. Ma, X. Zhang, H. T. Zheng, et al., “ Shufflenet v2: Practical Guidelines for Efficient Cnn Architecture Design,” in Proceedings of the European Conference on Computer Vision (ECCV, 2018), 116-131.

[42]

M. Tan and Q. Le, “ Efficientnet: Rethinking Model Scaling for Convolutional Neural Networks,” in International Conference on Machine Learning (PMLR, 2019), 6105-6114.

[43]

H. Zhang, C. Wu, Z. Zhang, et al., “ Resnest: Split-Attention Networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022), 2736-2746.

[44]

X. Ding, X. Zhang, N. Ma, et al., “ Repvgg: Making vgg-style Convnets Great Again,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2021), 13733-13742.

[45]

R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “ Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization,” International Journal of Computer Vision 128, no. 2 (2020): 336-359, https://doi.org/10.1007/s11263-019-01228-7.

[46]

Z. Yi-Kang, Z. Heng, L. Yong-Ge, et al., “ Oracle Character Recognition Based on Cross-Modal Deep Metric Learning,” Acta Automatica Sinica 47, no. 4 (2021): 791-800, https://doi.org/10.16383/j.aas.c200443.

[47]

L. Dazheng, “ Random Polygon Cover for Oracle Bone Character Recognition,” in Proceedings of the 2021 5th International Conference on Computer Science and Artificial Intelligence (2021), 138-142.

[48]

J. Li, Q. F. Wang, R. Zhang, et al., “ Mix-up Augmentation for Oracle Character Recognition With Imbalanced Data Distribution,” in International Conference on Document Analysis and Recognition (Springer International Publishing, 2021), 237-251.

[49]

Y. Li, T. M. Hospedales, Y. Z. Song, and S. Gong, “ Free-Hand Sketch Recognition by Multi-Kernel Feature Learning,” Computer Vision and Image Understanding 137 (2015): 1-11, https://doi.org/10.1016/j.cviu.2015.02.003.

[50]

S. Hayat, S. Kun, S. Shahzad, P. Suwansrikham, M. Mateen, and Y. Yu, “ Entropy Information-Based Heterogeneous Deep Selective Fused Features Using Deep Convolutional Neural Network for Sketch Recognition,” IET Computer Vision 15, no. 3 (2021): 165-180, https://doi.org/10.1049/cvi2.12019.

[51]

L. Zhang, “ Hand-Drawn Sketch Recognition With a Double-Channel Convolutional Neural Network,” EURASIP Journal on Applied Signal Processing 2021, no. 1 (2021): 1-12, https://doi.org/10.1186/s13634-021-00752-4.

[52]

L. Wang, S. Zhang, H. He, X. Zhang, and Y. Sang, “ A Hierarchical Residual Network With Compact Triplet-Center Loss for Sketch Recognition,” Multimedia Tools and Applications 81, no. 11 (2022): 15879-15899, https://doi.org/10.1007/s11042-022-12431-z.

[53]

S. Mishra, A. Shah, A. Bansal, et al., “ Learning Visual Representations for Transfer Learning by Suppressing Texture,” in 33rd British Machine Vision Conference Proceedings (BMVC, 2022).

[54]

C. Herrmann, K. Sargent, L. Jiang, et al., “ Pyramid Adversarial Training Improves Vit Performance,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022), 13419-13429.

[55]

X. Mao, G. Qi, Y. Chen, et al., “ Towards Robust Vision Transformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022), 12042-12051.

[56]

Q. Deng, S. Niu, R. Zhang, et al., “ Learning to Generate Gradients for Test-Time Adaptation via Test-Time Training Layers,” Proceedings of the AAAI Conference on Artificial Intelligence 39, no. 15 (2025): 16235-16243, https://doi.org/10.1609/aaai.v39i15.33783.

[57]

A. Kumar, A. Raghunathan, J. Robbie, et al., “ Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution,” in 10th International Conference on Learning Representations (ICLR, 2022).

[58]

Z. Xu, D. Liu, J. Yang, et al., “ Robust and Generalizable Visual Representation Learning via Random Convolutions,” arXiv preprint, arXiv:2007.13003, July 25, 2020, https://doi.org/10.48550/arXiv.2007.13003.

[59]

X. Li, X. Fan, J. Yu, Z. Dong, X. Cairang, and P. Lan, “ Jiu Fusion Artificial Intelligence (JFA): A Two-Stage Reinforcement Learning Model With Hierarchical Neural Networks and Human Knowledge for Tibetan Jiu Chess,” Frontiers of Information Technology & Electronic Engineering 26, no. 10 (2025): 1969-1983, https://doi.org/10.1631/fitee.2500287.

PDF (3054KB)

0

Accesses

0

Citation

Detail

Sections
Recommended

/

〈 〉