I am a recent graduate of the Master’s program in Electrical Computer Engineering at UC San Diego. Aspiring to create my personal portfolio, I have set up this platform to share ideas and express myself. Currently, I have returned to Hong Kong after leaving San Diego:)
In the past, I used to gravitate towards big cities, following trends and being influenced by the people around me. However, I had a realization about the side effects of metropolitan living, such as pollution and limited living space. This led me to explore more possibilities in life, including studying a new major and traveling to new countries. Now, I consider myself an adventurer and trailblazer, and I believe this attribute will greatly enrich my life.
After experiencing numerous unforeseen incidents in recent months, I began to comprehend the sense of powerlessness that individuals face and the importance of embracing uncertainty.
My interests are quite diverse, ranging from sports to music. On this website, I have documented some of my proudest moments, which you can access below…
My research interest includes computer vision and self-supervised learning. I have several related projects and academic journals with total google scholar citations 64+ (You can also use google scholar badge 
🔥 News
- 2023.11: 🎉🎉 I will start a new life in Beijing soon!
- 2023.06: 🎉🎉 I graduated from UCSD and returned to Hong Kong!
🧑💻 Blogs
Personal Blog with Sharings and Thoughts
📖 Educations
- 2021.09 - 2023.06, Master, University of California, San Diego, United States
- 2016.08 - 2020.06, Bachelor, The Hong Kong Polytechnic University, Hong Kong
- 2018.09 - 2019.01, Exchange Program, Cardiff University, United Kingdom
💻 Work Experience
- 2023.12 - Now, Computer Vision Algorithm Engineer,HAOMO.AI, Beijing, China
- 2022.06 - 2022.9, Data Scientist, TCL Corporate Research, Hong Kong
- 2020.07 - 2021.7, Research Assistant, Hong Kong Polytechnic University, Hong Kong
🎓 Projects
- Implemented a conditional diffusion model based on the SR3 model to generate high-resolution images from low-resolution inputs.
- Modified the training objective to predict residual between output and input, leading to faster convergence and improved preservation of details.
- Infused low-resolution inputs into bottleneck layers to enhance output consistency, resulting in improvements of PSNR (23.1) and SSIM (0.69).
Network Systems Project - Developed a web server for data retrieval using a custom HTTP/1.1 protocol which is robust to invalid request. Built a cloud‐based file storage service with metadata and data block server based on Dropbox, support concurrent data access by gRPC.
- Implemented a distributed data blocks storage system with consistent hashing algorithm for efficient and fault‐tolerant data storage.
- Replicated metadata server for fault‐tolerance and consistency, using the RAFT consensus protocol for consistent state machine updates.
Deep Learning for Image Captioning on Fliker 8K Dataset - Constructed an Encoder-Decoder structure with ResNet and LSTM respectively, and trained the model sequentially to generate captions with a unigram BLEU score of 0.28.
- Improved the caption generation ability by introducing the soft attention mask and achieved a BLEU score of 0.26.
- Further enhanced the model by replacing the encoder with a pre-trained Vision Transformer and the decoder with GPT-2, resulting in a total BLEU score of 0.53.
Find more projects at my github page.🎖 Honors and Awards
- 2018-2019, Dean’s List, Hong Kong Polytechnic University, Hong Kong
📝 Publications
400 MHz ultrafast optical coherence tomography
Dongmei Huang, Feng Li, Zijian He, Zihao Cheng, Chao Shang, PKA Wai
Ultrasensitive optofluidic enzyme-linked immunosorbent assay by on-chip integrated polymer whispering-gallery-mode microlaser sensors
Xia Ouyang, Tong Liu, Yangxi Zhang, Jijun He, Zijian He, A Ping Zhang, Hwa-Yaw Tam
Multifunctional textile platform for fiber optic wearable temperature-monitoring application
Ziyang Xiang, Liuwei Wan, Zidan Gong, Zhuxin Zhou, Zhengyi Ma, Xia OuYang, Zijian He, Chi Chiu Chan