Yifan LIU

No.28, Nanli Road · Hongshan District, Wuchang · Wuhan · Hubei Province · P. R. China · liu_yi_fan@hust.edu.cn

I graduated with honors in June 2025 from the Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, with a Ph.D. in Computer Science and Technology. My research focuses on low-level vision tasks and 3D reconstruction. I previously interned at the PKU-Wuhan Institute for Artificial Intelligence, Peking University. I am currently a lecturer at the School of Electrical and Electronic Engineering, Hubei University of Technology.

Experience

Lecturer

School of Electrical and Electronic Engineering, Hubei University of Technology

Specialized in city-scale 3D reconstruction from satellite imagery, with experience in end-to-end model development for urban modeling and digital twin applications. Contributed to the design of learning-based reconstruction pipelines and led the creation of large-scale datasets to support uncalibrated multi-view 3D generation.

July 2025 - present

Research intern

PKU-Wuhan Institute for Artificial Intelligence, Peking University

Developed a multi-view satellite image generation and alignment framework based on geolocation ranges for high-precision Level-3 3D mesh reconstruction. Designed a differentiable mesh generation method for multi-view 3D modeling tasks. Contributed to the application of large language models in transportation systems, including fine-tuning open-source LLMs and deploying them within Wuhan Integrated Transport Operation Center.

July 2024 - January 2025

Ph.D. in Computer Science and Technology

Huazhong University of Science and Technology

Specialized in atmospheric degraded image restoration, with extensive experience leading full-stack research projects involving data security, fault prediction, and cross-platform deployment. Proficient in algorithm design using PCA, AdaBoost, SVM, and Random Forest; developed encryption schemes, object detection networks, and managed end-to-end workflows on both Ubuntu and Kylin systems with ARM-based servers.

September 2021 - June 2025

papers

  1. MFID-Net: Multi-Scaled Feature-Fused Image Dehazing via Dynamic Weights [paper] [code]
  2. MOONLIT: momentum-contrast and large-kernel for multi-fine-grained deraining [paper] [code]
  3. Snowed autoencoders are efficient snow removers [paper] [code]
  4. You cannot handle the weather: Progressive amplified adverse-weather-gradient projection adversarial attack [paper] [code]
  5. Underwater vision clearing with latent space alignment for marine scene enhancement [paper] [code]
  6. Multi-Modal Image Reflection Removal with Prior Knowledge of Reflection Structure Inconsistency [paper 🔒] [code]
    🔐 Username: hpccpub24   Password: conf24//
  7. Sat2Mesh: Satellite-to-Mesh Urban Modeling from Multi-view Satellite Images for City-scale 3D Reconstruction [paper (under review)] [code] [demo]
  8. Detection of Insulator Defects With Improved ResNeSt and Region Proposal Network [paper]
  9. SKEAFN: sentiment knowledge enhanced attention fusion network for multimodal sentiment analysis [paper] [code]
  10. Analysis and Compensation of Installation Perpendicularity Error in Unmanned Surface Vehicle Electro-Optical Devices by Using Sea-Sky Line Images [paper]
  11. Enhancing Speech Emotion Recognition with Speech Dynamic Modeling and Multi-Modal Knowledge Distillation [paper]
  12. Multi-level Multi-task representation learning with adaptive fusion for multimodal sentiment analysis [paper]
  13. Insulator Defect Detection Based on EfficientDet and Binocular Camera [paper]
  14. A Pedestrian Motion Prediction Method Based on Improved Two-step Framework [paper]
  15. Insulator Identification Based on Improved Xception Method [paper]
  16. A Multi Focus Image Fusion Method Based on Improved Focus Region Detection [paper]
  17. Non-invasive Industrial Load Identification Based on KM-LSTM-AE Method [paper]
  18. PCB Board Defect Detection Based on Lightweight Artificial Neural Network [paper]
  19. Surface Defect Detection of PCB Based on Improved YOLOv5 [paper]
  20. Insulator defect detection method based on improved YOLOv5 neural network [paper]
  21. A method for detecting surface defects of solar cells based on improved YOLOv5 [paper]
  22. Surface Defect Detection Method of Solar Cell Based on Vision [paper]

Interests

My research interests span low-level vision and 3D modeling, including image restoration under challenging conditions such as haze, rain, snow, reflection, and underwater distortion. I also explore adversarial attacks in vision systems, multi-view 3D reconstruction, and multimodal large language models.

Technically, I'm proficient in Python, MATLAB, and C/C++, and experienced with PyTorch, Git, Linux/Windows environments, and common scientific computing tools such as NumPy and Selenium. I actively work on geometric perception and vision tasks requiring robustness and cross-domain generalization. Languages: Mandarin (native), English (fluent), Japanese (intermediate).

🕹️ When I'm not wrangling neural networks or taming satellite images into 3D meshes, I'm probably stuck somewhere between Lordran and Night City. A lifelong anime and game enthusiast, I find inspiration in the bizarre poses of JoJo's Bizarre Adventure, the intensity of Attack on Titan, and the glorious chaos of Chainsaw Man and Cyberpunk: Edgerunners.

🎮 From World of Warcraft raids to dying repeatedly in Elden Ring, Dark Souls, and Sekiro (on purpose, of course), I embrace the challenge. You'll also find me exploring dungeons in The Binding of Isaac, mining in Terraria, causing mayhem in GTA, or being a lone wolf in The Witcher and Fallout.

💍 Also, eternal thanks to my wife Jiayi Sun Jiayi Sun — my favorite player two in this wild open-world adventure.