Current Analyst @Morgan Stanley. Prev Quant Research Intern @y-intercept, Summer Analyst @Morgan Stanley, Data Science Intern @Societe Generale, Student Ambassador @Microsoft. CS '22 @HKU.
Fixels
Description
Fixels is a platform to help opthalmologists in generating quick, accurate, and reliable diabetic retinopathy diagnosis via retinal images. In Fixels, we use a convolutional neural network model trained on thousands of labeled retinal images from kaggle public datasets and messidor private datasets.
The model was trained on HKU GPU server, and yields an accuracy near to 80% on validation data. Finally, We connected the core functionality of Fixels to a web application as an interface which can be easily accessed and used by everyone.
Relevant Links
Tech and tools
- Frontend using React, written on TypeScript
- Model was built and trained on top of Keras APIs, written on Python, following object-oriented practice
- Flask as a RESTful API that connects both ends of the application