Jiangquan Hou

Jiangquan Hou

Undergraduate Student, Xidian University

I am an undergraduate student majoring in Electronic Information Engineering at Xidian University. My research interests include image multimodal fusion, semantic segmentation, multimodal large models, and incremental learning.

🔥 News

  • 2026.07:  🎉 Started as a Research Assistant at The Chinese University of Hong Kong, Shenzhen.
  • 2025.09:  Began research on rehearsal-free class-incremental learning.
  • 2025.04:  Started the HiMoE project on multimodal image segmentation.

📝 Publications

IEEE TIP (under review)
sym

HiMoE: Hierarchical Mixture-of-Experts Dynamic Fusion for Multimodal Image Segmentation

Xueli Geng, Jiangquan Hou, Lingling Li, Licheng Jiao, Xu Liu, Fang Liu, Wenping Ma

Manuscript under review at IEEE Transactions on Image Processing

  • Designed a Gated ConvMamba Block hybrid encoder integrating CNN local feature extraction and Mamba long-range dependency modeling.
  • Proposed a Spatial-Frequency Fusion (SFFusion) module combining spatial saliency and frequency spectral features for adaptive cross-modal feature fusion.
  • Built a HiMoE decoder with hierarchical pixel-wise gating, validated against 10+ SOTA methods on 4 remote sensing benchmarks.

🔬 Research Experience

Missing-Modality-Aware Visual Question Answering for Remote Sensing Multimodal Large Models 07/2026 - Present · Research Assistant, The Chinese University of Hong Kong, Shenzhen Supervised by Prof. Qingyu Li

  • Developed a programmatic pipeline for constructing a multimodal remote sensing VQA benchmark with missing-modality configurations and rule-based quality control.
  • Designed missing-modality-aware modules to improve model robustness under incomplete multimodal inputs.
  • Developed a fine-tuning and evaluation pipeline for adapting Qwen3-VL under complete- and missing-modality settings.

Subspace-aware Angular-margin Gaussian Ensemble for Rehearsal-Free Class-Incremental Learning 09/2025 - Present · Xidian University Supervised by Prof. Tianyang Zhang

  • Proposed a rehearsal-free class-incremental learning framework on frozen ViT using task-specific adapters in dedicated subspaces for knowledge isolation.
  • Designed a sine-guided angular margin optimization with adaptive hard-negative constraints for high-similarity remote sensing categories.
  • Introduced a variance-weighted Gaussian inference module with per-subspace Z-score calibration for reliable cross-task prediction.

HiMoE: Hierarchical Mixture-of-Experts Dynamic Fusion for Multimodal Image Segmentation 04/2025 - 03/2026 · Xidian University Supervised by Prof. Xueli Geng

  • Proposed the HiMoE dynamic fusion framework achieving SOTA segmentation performance with only 2.18M parameters.
  • Designed a Gated ConvMamba Block hybrid encoder integrating CNN local feature extraction and Mamba long-range dependency modeling.
  • Proposed a Spatial-Frequency Fusion (SFFusion) module combining spatial saliency and frequency spectral features for adaptive cross-modal fusion.
  • Built a HiMoE decoder with hierarchical pixel-wise gating, validated against 10+ SOTA methods on 4 remote sensing benchmarks.

Research on Human Action Recognition Integrating Visual SLAM and Dual-Stream CNN 12/2024 - 01/2025 · NC State University Supervised by Prof. Kevin Han

  • Proposed a vision-based solution leveraging Visual SLAM and CNN to identify unsafe worker behaviors from helmet-mounted cameras.
  • Designed a two-stream 1-D CNN model for body action recognition and head gesture recognition.
  • Achieved over 80% accuracy with real-time performance (0.56s average latency).

💻 Projects

  • deep-explainer - Turn any academic paper or GitHub repo into a beautiful HTML explainer page with KaTeX, Mermaid, dark mode, and example-driven explanations.
  • agnescode - Terminal-native AI coding agent with built-in Agnes AI (free multimodal API).
  • Token-Monitor - Floating AI token usage monitor with a collapsible ball, real-time tracking of API costs and quota. Built with Tauri.

🎖 Honors and Awards

  • 2025 Second Prize (Shaanxi Province), National Undergraduate Mathematical Contest in Modeling
  • Xidian University First Class Scholarship
  • China-France Cooperative Education Endeavor Scholarship, Xidian University

📖 Education

  • 09/2023 - 07/2027, B.E. in Electronic Information Engineering, Xidian University, Xi’an, China.

🏃 Extracurricular Activities

  • 03/2025 - 09/2025, Head of the Coordination Department, Student Union of School of Electronic Engineering, Xidian University.
  • 01/2024 - Present, Deputy Captain of the Admissions General Team / Recruitment Ambassador for South China, Xidian University. Awarded Outstanding Admission Worker.
  • 01/2024 - 02/2024, Team Leader, Winter Vacation School Visit Campaign. Awarded Outstanding Individual and Outstanding Team Leader.