Selected work

Representative Papers

NeuroAdaptTrainer: A Fiji/ImageJ Plugin for YOLO-Based Neuron Segmentation, InteractiveCorrection and Transfer Learning

Aug 05, 2026

This work addresses the inefficiency and reliance on manual labor in neuronal segmentation and counting within neuroscience research by developing a Fiji/ImageJ plugin that integrates a YOLO-based instance segmentation model. The tool enables fully automated neuron detection, supports interactive manual correction, and incorporates online transfer learning using user-provided annotations to continuously refine the model. It also includes a built-in validation module for quantitative performance evaluation. By seamlessly embedding deep learning into established microscopy image analysis workflows, this approach significantly lowers the barrier for non-expert users to leverage state-of-the-art segmentation models while preserving expert oversight, thereby enhancing both segmentation efficiency and model adaptability.

0 citationsRead paper

A Self-Supervised Approach for Enhanced Feature Representations in Object Detection Tasks

Feb 18, 2026

This work proposes a self-supervised feature learning method specifically designed for object detection to address the heavy reliance on large-scale annotated data. By pretraining the feature extractor on unlabeled data and guiding the model to focus on semantically informative object regions, the approach significantly enhances the representational capacity of the detector under limited annotation budgets. Experimental results demonstrate that the proposed method outperforms conventional ImageNet-pretrained models across multiple object detection benchmarks, achieving not only improved detection accuracy but also greater robustness and reliability.

0 citationsRead paper

A review on data fusion in multimodal learning analytics and educational data mining

Nov 25, 2025

This study addresses the challenges of fusing heterogeneous educational data—such as audio, video, eye-tracking, physiological signals, and behavioral logs—in multimodal learning analytics (MLA), and the consequent lack of robust intervention support. We systematically review and establish a taxonomy and technical pathway for data fusion in educational contexts. Methodologically, we propose a novel three-tier fusion framework spanning feature-level, decision-level, and model-level integration, synergizing machine learning and educational data mining techniques to enhance cross-modal collaborative modeling. Our analysis identifies critical bottlenecks in temporal alignment, interpretability, and real-time intervention capability, and clarifies a theoretical paradigm and developmental roadmap for data fusion in intelligent learning environments. Results demonstrate that principled multimodal fusion significantly improves learning state recognition accuracy and the efficacy of pedagogical interventions, thereby providing a methodological foundation for next-generation adaptive learning systems.

0 citationsRead paper

A Multimodal Conversational Assistant for the Characterization of Agricultural Plots from Geospatial Open Data

Sep 22, 2025

Non-expert users face significant challenges in efficiently leveraging open Earth observation (EO) and agricultural remote sensing data for sustainable land management. Method: We propose an open-source multimodal conversational assistant built upon a retrieval-augmented generation (RAG) framework that integrates orthoimagery, Sentinel-2 vegetation indices, and textual agricultural documentation. The system employs Qwen3-32B as its foundation language model to enable zero-shot natural language interaction. Crucially, we introduce an LLM-as-a-judge unsupervised, multi-dimensional automated evaluation mechanism to enhance reproducibility and cross-regional generalizability. Contribution/Results: Experiments demonstrate that the assistant generates accurate, context-aware, and interpretable responses characterizing individual farmland parcels. Comprehensive evaluation across multiple dimensions confirms the system’s effectiveness, robustness, and scalability. By abstracting technical complexity, it substantially lowers the barrier to entry for agricultural remote sensing data utilization among non-specialist stakeholders.

0 citationsRead paper

Remote sensing colour image semantic segmentation of trails created by large herbivorous Mammals

Apr 16, 2025

This study addresses the challenge of semantic segmentation of grazing trails—long, thin, low-contrast structures in remote-sensing RGB imagery—whose delineation is severely hindered by background clutter and poor contrast. We propose UNet-MambaOut, a novel architecture integrating MambaOut as the encoder (to capture long-range temporal dependencies) and UNet as the decoder (to recover fine-grained spatial details). To our knowledge, this is the first end-to-end deep learning framework achieving high-accuracy semantic segmentation for this task. Evaluated on multi-scene aerial imagery, it significantly improves both accuracy and structural integrity in delineating bare-soil grazing trails. Our method is the first globally to attain competitive segmentation performance (mIoU > 72%) for this specific application. It enables precise identification of biodiversity-sensitive areas, thereby supporting habitat dynamics monitoring and adaptive land management.

0 citationsRead paper

Recent publications

Latest Papers

NeuroAdaptTrainer: A Fiji/ImageJ Plugin for YOLO-Based Neuron Segmentation, InteractiveCorrection and Transfer Learning

Aug 05, 2026

This work addresses the inefficiency and reliance on manual labor in neuronal segmentation and counting within neuroscience research by developing a Fiji/ImageJ plugin that integrates a YOLO-based instance segmentation model. The tool enables fully automated neuron detection, supports interactive manual correction, and incorporates online transfer learning using user-provided annotations to continuously refine the model. It also includes a built-in validation module for quantitative performance evaluation. By seamlessly embedding deep learning into established microscopy image analysis workflows, this approach significantly lowers the barrier for non-expert users to leverage state-of-the-art segmentation models while preserving expert oversight, thereby enhancing both segmentation efficiency and model adaptability.

0 citationsRead paper

A Self-Supervised Approach for Enhanced Feature Representations in Object Detection Tasks

Feb 18, 2026

This work proposes a self-supervised feature learning method specifically designed for object detection to address the heavy reliance on large-scale annotated data. By pretraining the feature extractor on unlabeled data and guiding the model to focus on semantically informative object regions, the approach significantly enhances the representational capacity of the detector under limited annotation budgets. Experimental results demonstrate that the proposed method outperforms conventional ImageNet-pretrained models across multiple object detection benchmarks, achieving not only improved detection accuracy but also greater robustness and reliability.

0 citationsRead paper

A review on data fusion in multimodal learning analytics and educational data mining

Nov 25, 2025

This study addresses the challenges of fusing heterogeneous educational data—such as audio, video, eye-tracking, physiological signals, and behavioral logs—in multimodal learning analytics (MLA), and the consequent lack of robust intervention support. We systematically review and establish a taxonomy and technical pathway for data fusion in educational contexts. Methodologically, we propose a novel three-tier fusion framework spanning feature-level, decision-level, and model-level integration, synergizing machine learning and educational data mining techniques to enhance cross-modal collaborative modeling. Our analysis identifies critical bottlenecks in temporal alignment, interpretability, and real-time intervention capability, and clarifies a theoretical paradigm and developmental roadmap for data fusion in intelligent learning environments. Results demonstrate that principled multimodal fusion significantly improves learning state recognition accuracy and the efficacy of pedagogical interventions, thereby providing a methodological foundation for next-generation adaptive learning systems.

0 citationsRead paper

A Multimodal Conversational Assistant for the Characterization of Agricultural Plots from Geospatial Open Data

Sep 22, 2025

Non-expert users face significant challenges in efficiently leveraging open Earth observation (EO) and agricultural remote sensing data for sustainable land management. Method: We propose an open-source multimodal conversational assistant built upon a retrieval-augmented generation (RAG) framework that integrates orthoimagery, Sentinel-2 vegetation indices, and textual agricultural documentation. The system employs Qwen3-32B as its foundation language model to enable zero-shot natural language interaction. Crucially, we introduce an LLM-as-a-judge unsupervised, multi-dimensional automated evaluation mechanism to enhance reproducibility and cross-regional generalizability. Contribution/Results: Experiments demonstrate that the assistant generates accurate, context-aware, and interpretable responses characterizing individual farmland parcels. Comprehensive evaluation across multiple dimensions confirms the system’s effectiveness, robustness, and scalability. By abstracting technical complexity, it substantially lowers the barrier to entry for agricultural remote sensing data utilization among non-specialist stakeholders.

0 citationsRead paper

Remote sensing colour image semantic segmentation of trails created by large herbivorous Mammals

Apr 16, 2025

This study addresses the challenge of semantic segmentation of grazing trails—long, thin, low-contrast structures in remote-sensing RGB imagery—whose delineation is severely hindered by background clutter and poor contrast. We propose UNet-MambaOut, a novel architecture integrating MambaOut as the encoder (to capture long-range temporal dependencies) and UNet as the decoder (to recover fine-grained spatial details). To our knowledge, this is the first end-to-end deep learning framework achieving high-accuracy semantic segmentation for this task. Evaluated on multi-scene aerial imagery, it significantly improves both accuracy and structural integrity in delineating bare-soil grazing trails. Our method is the first globally to attain competitive segmentation performance (mIoU > 72%) for this specific application. It enables precise identification of biodiversity-sensitive areas, thereby supporting habitat dynamics monitoring and adaptive land management.

0 citationsRead paper