Han Yi

Hi! I’m currently a research intern at the National University of Singapore (NUS), NExT++ Research Center, advised by Prof. Tat-Seng Chua, Dr. Zhedong Zheng, and Prof. Xiangyu Xu.

Open-source projects can be found at my [Github], and publications can be found at ORCID.

More details can be found in my [CV].

Research Statement

I am interested in research related to deep learning and computer vision. My work focuses on NeRF, Text-to-3D generation and editing.

Publications

1. Progressive Text-to-3D Generation for Automatic 3D Prototyping.
   • Authors: Han Yi, Zhedong Zheng, Xiangyu Xu, Tat-seng Chua
   • Conference: IEEE Conference on Robotics and Automation (ICRA 2024) (Submitted)
2. Image Deblurring with Image Blurring.
   • Authors: Ziyao Li, Zhi Gao, Han Yi, Yu Fu, Boan Chen
   • Journal: IEEE Transactions on Image Processing (TIP 2023)
3. Single Image Deraindrop Leveraging Luminance Priors and Context Aggregation.
   • Authors: Yi Liu, Zhi Gao, Tiancan Mei, Han Yi
   • Journal: IEEE Transactions on Image Processing (TIP 2023) (Under review)
4. Motion Blur Synthesis for Image Deblurring via Disentangling Latent Blur Space.
   • Authors: Ziyao Li, Han Yi, Zhiyu Zhou, Xuhui Zhao, Zhi Gao
   • Journal: IEEE Signal Processing Letters (SPL 2023) (Under review)
5. A hierarchical geometry-to-semantic fusion GNN framework for earth surface anomalies detection.
   • Authors: Boan Chen, Aohan Hu, Mengjie Xie, Zhi Gao, Xuhui Zhao, Han Yi
   • Conference: International Conference On Brain-Inspired Cognitive Systems (BICS 2023) (Best Student Paper)
6. How Challenging is a Challenge for SLAM? An Answer from Quantitative Visual Evaluation.
   • Authors: Xuhui Zhao, Zhi Gao, Hao Li, Chenyang Li, Jingwei Chen, Han Yi
   • Conference: International Conference On Brain-Inspired Cognitive Systems (BICS 2023)

Research experience

1. Progressive Text-to-3D Generation for Automatic 3D Prototyping

   • Duration: Dec. 2022 - Present
   • Institution: NExT++ Research Center, National University of Singapore
   • Advisor: Prof. Tat-seng Chua
     • Proposed a Multi-Scale Triplane Network (MTN) to gradually create the 3D model in a bottom-up style.
     • Proposed a progressive learning strategy to refine details of the 3D model.
     • Achieved high-resolution outputs that align closely with natural language descriptions.

2. Image Deblurring with Image Blurring

   • Duration: Nov. 2021 - Oct. 2022
   • Institution: Wuhan University
   • Advisor: Prof. Zhi Gao
     • Proposed a novel motion deblurring framework.
     • Synthesized a blur-sharp paired blur dataset Rear-Blur-COCOmini.
     • Obtained state-of-the-art deblurred results and introduced Variance of Laplacian edge detection (VL).

3. LED Visible Light Positioning Algorithm

   • Duration: July. 2021 - Sep. 2021
   • Institution: Tsinghua University
   • Advisor: Prof. Hao Zhang
     • Programmed LED lights on the ceiling with Manchester coding.
     • Utilized Solvepnp algorithm from OpenCV.
     • Improved the algorithm in various ways including special positions and turning angles.
     • Introduced Malformation Matrix to improve accuracy.

4. Maintenance System with Predictability Figuration

   • Duration: Sep. 2020 - Mar. 2021
   • Institution: Tsinghua University
   • Advisor: Prof. Hao Zhang
     • Built predictive models to calculate RUL (Remaining Useful Life).
     • Utilized Abrupt Detection, such as Pettitt Detection and Mann-Kendall Detection.
     • Optimized the efficiency with Cumulative Sum Control Chart and Exponentially Weighted Moving-Average.