About me
Hello! My name is Zhiyi Wang, and I am a junior student at ShanghaiTech University. My area of expertise lies in Computer Vision and Computer Graphics. Iām familar with contemporary Deep Learning frameworks such as MLP, CNN, RNN, AE, GAN, NeRF, Diffusion, and LLM related stuff, alongside rendering techniques like rasterization and surface/volume rendering. Currently, I am conducting research at SooLab lab, under the esteemed guidance of Prof. Sibei Yang. You can download my resume here.
Projects
Pixmath | Nov. 2022 ā Jan. 2023
A math formula recognizer designed as the final project of Deep Learning (CS280)
- Aiming to recognize math formulas in images rendered by LaTeX engine back to LaTeX code.
- Used a CNN-GRU pipeline with attention mechanism.
- Capable of attaining a 25% accuracy rate and a word error rate (WER) of 25% on the CROHME 2013 test set.
- Yields credible recognition outcomes for inline cases, although performance may be constrained by limited GPU resources.
Realistic Smoke Rendering | Dec. 2022 ā Jan. 2023
A renderer supports both surface and volume rendering designed as the final project of Computer Graphics (CS171)
- Implemented basic Monte-Carlo Path-Tracing for surface rendering.
- Implemented sampler with Delta-Tracking and transmittance estimation with Ratio-Tracking, in addition to their integration with surface rendering.
