About Me
I am an innovative deep worker, who consistently generates new ideas and commits to deep work to transform those ideas into tangible products. My superpower is turning creative and unconventional ideas into reality. Obtained both Bachelor's and Master's degree at UC Berkeley, I have worked with many brilliant people on class and research projects. As a team player, I might not always come up with the best idea, but I am always the one who solidifies all the details of our big visions and writes the code to make them a reality.
- Deep Learning
- Computer Vision
- Software Development
- Parallel/High-PerformanceComputing
My Projects
The City of Pixels
A pipeline for reconstructing large scale city scenes as 3D point clouds using RGB and depth data from the Google Street View API. The inputs used are 360 panoramas of street views and their associated depth map. The resolution is 256 by 512. The pipeline takes in the inputs, renders 3D point clouds, and aligns them to form a large 3D scene.
Vision-Language Model for Pose Estimation
Fine-tune the BLIP vision-language model on MPII Pose Estimation dataset. BLIP was pre-trained on image captioning. The fine-tuned model outputs captions with coordinates describing the location of each body joint. The resulting model is robust in scenes with physical occlusions and achieves a validation accuracy of 90%.
Autonomous Senior Helper System
A computer-vision-based fall detection system and can detect human falls regardless of any occlusions in the scene. Spatial-Temporal Neural Networks with Learnable Edges (STGCN-LE) is used as the backbone model and AlphaPose is used to extract skeleton data from each video frame for inference.
Let's Connect
I'am currently looking for new opportunities, my inbox is always open. Whether you have a question or just want to say hi, I will get back to you as soon as possible!
