Author name cluster

Dakuo He

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

1 author row

EAAI Journal 2026 Journal Article

Maintaining the consistency of small targets on invariant deep semantic structures

Haifeng Sang
Yuwei Wu
Qing Liu
Chenxin Liu
Xinyan Chang
Dakuo He

In Infrared Small Target Detection (IRSTD), weak target signals and low contrast make the boundaries of small targets difficult to distinguish from complex backgrounds. The multi-level downsampling in the encoder further attenuates boundary information, while upsampling and cross-layer fusion in the decoder may amplify residual noise and pseudo-edge responses. The combination of these effects poses significant challenges for accurate boundary reconstruction and semantic discrimination. To address this issue, we propose the Edge-Target Deep Semantic Consistency (ET-DSC) semantic adaptive balancing framework: in encoder, shallow-layer modeling and gated fusion are adopted to enhance target boundaries; in decoder, semantic consistency constraints are introduced to preserve real boundaries and suppress false edges. Furthermore, a semantic allocation mechanism is established between shallow and deep layers to achieve cooperative optimization between edge compensation and semantic preservation. Experimental results on multiple public IRSTD datasets demonstrate that ET-DSC effectively reconstructs small-target boundaries and achieves higher localization accuracy under complex and low Signal-to-Noise Ratio(SNR) conditions. This work provides a reliable framework for fine-grained modeling of small targets in infrared scenes and offers new insights for future IRSTD network design. The codes are available at https: //github. com/Yuweiw-1024/ET-DSC.

Details DOI

EAAI Journal 2025 Journal Article

A meta-contrastive learning hybrid model for adaptive temperature trend prediction in variable ladle preheating

Youcheng Zong
Runda Jia
Shuai Wu
Liqiang Zhang
Dakuo He

Predicting the future temperature trends during ladle preheating is crucial for enhancing production efficiency and ensuring molten steel quality. However, due to the complexity of the preheating process, uncertainties in the external environment, and irregular preheating operations, temperature trends become difficult to predict. To address these problems, a meta-contrastive learning-based hybrid model (MCLHM) is proposed in this paper. It achieves multi-task differentiation and adapts to external environmental changes by employing a meta-contrastive learning strategy. The hybrid model consists of an environment-aware submodel, a temperature prediction main model, and a recursive output module, enabling multi-step confidence interval predictions with temporal correlations. The efficiency of the proposed method is enhanced by combining data-driven approaches with thermodynamic mechanisms. Field data experiments demonstrate that MCLHM outperforms existing methods. This method has been deployed in industrial settings, verifying its feasibility.

Details DOI

EAAI Journal 2023 Journal Article

Decision system for copper flotation backbone process

Haipei Dong
Fuli Wang
Dakuo He
Yan Liu

This study proposed a decision system that can output the flotation backbone flowchart using the natural properties of copper ore. The proposed decision system includes three decision tasks: product scheme, flotation scheme and grinding scheme. Each decision task is a multi-label classification problem. To improve the classification effect of each sub-label, extreme gradient boosting (XGBoost) is used as a subclassifier, because of its ability to deal with small and high-dimensional samples. To selectively utilize the relations between the sub-labels in the same task, a modified classifier chain (MCC) was proposed. To specifically use the effect of a front-end task on a back-end task, the decision system connects the MCC-XGBoost corresponding to the three tasks in series. Accordingly, the outputs of a front-end task becomes the candidate features of a back-end task. To improve the recall rates of minority classes, the classification thresholds are customized using the Yoden index. Finally, the high performance of the decision system was demonstrated by hypothesis testing.

Details DOI