Medical diagnosis should comprehensively consider the patient's medical history and background. These diagnoses are based on clinical results and theoretical grounds reported in academic medical reports. This project aims to build a system to assist individuals in performing preliminary medical diagnoses. By using RAG (Retrieval-Augmented Generation) and supplementing it with information from LLM (Large Language Model) engines and search engines, the system can reflect a broader range of information for more accurate medical diagnosis decisions. 

YOLY(You Only Look at Yourself) is a real-time visual and haptic guidance system for golf swing in virtual reality. Users can correct their posture according to the first-person and third-person perspectives displayed on the virtual screen seen through the Head-Mounted Display (HMD). The virtual screen corresponds with the user’s head movements, mitigating the “look-aways” phenomenon. Additionally, in accordance with related research, we divided the training sessions into a practice phase and a full-swing phase. The practice phase comprises seven stages: address, back-swing, top-of-swing, down-swing, impact, follow-through, and finish. We selected four steps before impact. Each phase gives the user visual guidance and haptic feedback in real-time.  

ZooBlox is an innovative project that integrates smart block toys with a tablet application powered by generative AI, aiming to revolutionize the way children engage with toys and technology. The project focuses on enhancing children’s creativity by enabling them to construct imaginary animals using physical blocks and digital technology. In this application, generative AI plays a pivotal role in transforming these creations into virtual beings within a digital zoo, encouraging imagination, confidence, and technological familiarity in children. This innovative approach, merging physical play with AI technology, promotes creativity and prepares children for future technological landscapes. 

Product Manager

This project is Avant Garde’s first PFP collection. A limited edition line of PFP NFTs inspired by the work of renowned Korean artist Jungwoo Ha. The collection was generated with more than 500 unique wearable items, each curated with Ha Jung Woo’s signature style. With only 978 NFTs available, this collection offers a rare opportunity to own a unique piece of art that has the touch of both Ha Jung Woo and SuperNormal. Avant Garde designs and produces digital fashion NFTs which can be worn in any type of digital world such as metaverses and online games. 

Team Lead

We quantified how healthy the community is, and presented it along with the price data of a specific project so that the data acts as a major factor in users’ investment. In addition, the {on, off} Chain data of the Crypto project was visually shown to help intuitive understanding. UX improvement was promoted by using a chatbot so that it could be viewed momentarily in the messenger. This project is implemented in Python and Dune Analytics. 

Team Lead

Project 1. A Person Re-Identification and Tracking Algorithm in Mobile Environment for Human-Robot Interaction

• Research details: The aim is to implement a system for tracking individuals in environments such as homes and offices by performing object detection and re-identification on mobile devices. The process involves choosing a lightweight network suitable for mobile environments and optimizing each model for the task at hand, allowing for high performance identification and tracking of registered users in real-time video.

Project 2. Lightening of the high-performance Pose Estimation model in the video using Pointwise Similarity

• Research details: The goal is to apply pose estimation to videos. The aim is to design a network that incorporates high accuracy and frame rate by utilizing the correlation between video frames, and to try different encoding methods and frequencies in order to achieve optimal results.