Projects | Dat NGUYEN

Computer Vision

Traffic Sign/Light Detection

One of the most challenging project I have worked on. The detection model is required to be robust to objects with different conditions like weather, daytime, lighttime, extreme light, etc. It is also mandatory to achieve optimal threshold in computation for deployment in AGX Driver or ARM64. The number of traffic signs and lights are around 300 classes.

Violence Detection

Designing a model that receives a video as input and predicts probability of sign of violence action.

Face Recognition

We developed a Face Recognition system that only require CPUs for training on a small-size training dataset, but can compete with SOTA in term of accuracy and speed. Moreover, the system has online learning ability. We published 2 papers with the system.

Recommendation System

Liquor Recommendation

This project I directly worked with our customer in Japan. My task is to develop a Recommendation Algorithm that recommends similar liquors based on the given taste coefficients. My algorithm finally show superior performance over the very well-known sake recommendation website (https://sakenowa.com/). I also published 2 prestigious papers with the project.

Web Development

Insight Data Science

In this project, I focus on the front-end site, working in ReactJS to display the content returned from the back-end APIs.

Collantotte

The first project I joined as a Web Developer after my bachelor graduation. I am in charge of building the admin site in which the administrator of the website can manage the selling products with easy-to-use, reliable, UI.

JAMJA

For this project, I am in the back-end site in which I wrote the APIs to return for the front-end. My mostly focused task is to use Elasticsearch to build a multi-level search bar with diverse creteria such as locations, closest distance, most-viewed, brands...