Implemented the UI of Windows app for Instance Segmentation task. The project, collaborated between UVL Lab and Hyundai Heavy Industries, aims to develop a human detection system through Hyundai factories’ cameras.  Worked on Hand Mesh Reconstruction model from 2D images of interacting hands. By manually annotating low- quality images and retraining, the model performs better performance than previous works.  [Python, C#, Pytorch, Numpy]

Ver más

Startup company with main product, Soundcorset, downloaded 30+ million times on Google Play. New application, ”Looina”, will be the first application where the type theory is extensively applied.  Designed and implemented front-end part and new semantics for core of Looina’s type system e.g. type inheritance, type dependency, comparator for type instances. Analyzed and refactored legacy code to improve maintainability.  Worked on identifying and resolving instances where type comparisons led to errors or inefficiencies. [JavaScript, TypeScript, Redom, CSS, HTML5, D3.js, Git]

Ver más

Classification of Astronomical Objects (Stars, Galaxies, Quasars.) using Machine Learning Models; Random Forest, and CNN. Worked on classifying more than 3 million spectroscopic/image data with up to 90% accuracy.  Weighing Black Holes with Deep Learning. Worked on Implementation of sophisticated Neural Network algorithm on more that 30 thousand spectroscopically confirmed quasars. [Python, Pytorch, Numpy, Deep Learning]

Ver más