Update: February 28th , 2025

Ongoing Project

Preparing for the Postgraduate Entrance Examination

I am currently in the process of preparing for the postgraduate entrance examination. I’m working hard on reviewing various professional knowledge, strengthening my English skills, and constantly doing a large number of practice questions to improve my problem-solving abilities. I’ve made a detailed study plan to ensure that I can systematically cover all the subjects and topics that will be tested, aiming to achieve good results in the upcoming exam.

(In progress)

iRetexturing: Intelligent Fashion Items Retexturing via Diffusion Models

• Supervised by Prof.Yao Jin and Dr.Xin Huang.

This paper introduces iRetexturing, a new framework for intelligent fashion items retexturing that integrates diffusion models to address the persistent challenge of preserving structural fidelity during texture transfer. While existing methods often compromise photorealistic quality or intuitive control, our quadripartite continuous repainting strategy synergizes texture expansion, cross-quadrant feature synchronization, and adaptive conditioning networks to achieve seamless material integration.

(SCI JRC:Q2 under review)

---

Early Project

Implementation of sketch-based algorithm for garment image generation

In an undergraduate mentor project, garment images were generated using garment texture patches and sketch outlines. Multiple papers such as styleGAN and other cutting edge deep learning models were called based on Linux during the thesis implementation to implement the data processing and some of the image algorithms.

Technology selection: Python, Pytorch, Linux, Deep Learning, Image Algorithms.

Undergraduate Mentorship Projects

Haier Home Appliances KV Products Indian Style Background Image Generation Workflow

This project aims to develop a set of industrial-grade standardized AI for Haier Home Appliances series product effect images for the Indian market Workflow. I completed data labeling for several home appliances and trained separate SDXL-based background lora.

I have independently developed the Haier home appliance workflow and home appliance set length drawing workflow, summarized the efficient project prompt template, and the straight-out KV drawing rate is as high as 75%.

In the second phase of the project, independently developed Flux-based man-machine light and shadow interaction workflow, with the ability to perfect the drawing.

Technology selection: ComfyUI, image algorithm, AIGC, Stable Diffusion, Flux.

Zhejiang University Artificial Intelligence Industrial Park Internship Program Hangzhou ZAOWUYUN Technology Co.