Summary
Overview
Work History
Education
Skills
Timeline
Generic
Seunghyeon Lee

Seunghyeon Lee

Ewha Womans University

Summary

I am an undergraduate student majoring in Artificial Intelligence at Ewha Womans University, with a strong interest in computer vision and generative AI.

Through coursework, I have developed proficiency in core AI technologies including Python and various machine learning and deep learning frameworks.

By participating in the ICP Lab internship and working on a diffusion steering project, I gained hands-on experience in the research process, including writing a paper.

Recently, I have developed a strong interest in 3D Gaussian Splatting and 3D-aware generative models, and I am eager to further explore research in 3D vision–based generative AI.

Overview

2026
2026
years of professional experience

Work History

Intern

ICP LAB in Ewha Womans University
12.2024 - Current
  • Participating in a research project on multi-view diffusion models, focusing on enhancing view consistency in text-to-image generation.
  • a paper submitted to the Broadcast and Media Engineering Conference (방송미디어공학회), presenting experimental findings on prompt-based consistency analysis using MVDream.
  • Attended the IPIU Conference, which led to a deepened interest in 3D Gaussian Splatting and 3D-aware generative models.



Project

  • Controlled Object Counting via Cross-Attention Steering in Diffusion ModelsApplied cross-attention steering techniques to guide object count accuracy in text-to-image diffusion generation.Evaluated on quantity-sensitive prompts (e.g., “three cats”, “five trees”) with improved alignment between textual count and image content.
  • View-Consistency Analysis of Multiview DiffusionAnalyzed view consistency in a Multiview Diffusion model (MVDream) with different prompts and submitted the results to the Broadcast and Media Engineering Conference (방송미디어공학회, 2025)
  • DeepFake Detection web – Developed an Android application for detecting deepfakes by combining face detection (MTCNN) and a CNN-based classifier (EfficientNet-B4). Integrated a user voting system to implement adaptive learning, enabling feedback-driven model refinement over time.
  • LLM Tuning (NLP) –Built a LangChain-based RAG system combining FAISS and Upstage LLM to answer user queries based on embedded document chunks. Included PDF parsing, chunking, vector embedding, and prompt-based generation using solar-1-mini-chat.
  • Multimodal Drunk Driving Detection – Built a multimodal system that detects drunk behavior by combining features from eye aspect ratio, skin tone, and voice. I was in charge of developing the audio model using paralinguistic features extracted via OpenSMILE. This project won the Gold Prize at the Ewha AI Challenge.
  • Time Deposit Subscription Prediction – Predicted bank term deposit subscriptions using an ensemble ML pipeline with CatBoost and XGBoost, including MICE imputation and outlier detection.

Awards

- Gold Prize, Ewha Womans University AI Competition“Multimodal Drunk Driving Detection System”

Education

Bachelor's Degree - Artificial Intelligence

Ewha Woman's University
Seoul
02-2026

Skills

Languages: Python, C, Java

Timeline

Intern

ICP LAB in Ewha Womans University
12.2024 - Current

Project

Awards

Bachelor's Degree - Artificial Intelligence

Ewha Woman's University
Seunghyeon LeeEwha Womans University