Hojun Choi

Aspiring Data Scientist CV ML Python

🛠 Projects

Here are some of the key projects I’ve worked on — from academic experiments to real-world applications.

🕺 Pose Sync Evaluator

  • Type: Pose Similarity Feedback System
  • Goal: Compare two short dance videos and provide basic feedback based on movement alignment
  • Tech: MediaPipe, OpenCV, fastdtw, ffmpeg
  • Highlights:
    • 2D Pose extraction from reference/user videos
    • DTW-based similarity scoring
    • Simple real-time feedback interface (planned)
  • Status: ✅ Completed
  • GitHub Repo

🚗 Seoul Smoking Zone GIS Map

  • Type: GIS + Urban Policy
  • Goal: Visualize legal smoking areas in Seoul (starting with Yongsan & Yeongdeungpo)
  • Tech: Pandas, Folium, Geocoding APIs
  • Highlights:
    • Address to coordinate conversion
    • Outlier removal & spatial clustering
  • Status: ✅ Completed
  • GitHub Repo

🍱 Food Image Classification

  • Type: Image Classification
  • Goal: Classify food images into predefined categories
  • Tech: PyTorch, CNN, Transfer Learning
  • Highlights:
    • Custom dataset crawling and augmentation
    • Model performance comparison (baseline vs. transfer learning)
  • Status: ✅ Completed
  • GitHub Repo

🏆 SCU AI Competition: Electricity Demand Forecasting (2025)

  • Type: Time Series Forecasting (Competition)
  • Goal: Predict future electricity demand with high accuracy
  • Tech: LightGBM, RandomForest, VotingClassifier
  • Highlights:
    • Hold-out validation + feature engineering
    • Model ensemble with hyperparameter tuning
  • Status: ✅ Completed
  • GitHub Repo

🧠 SCU AI Competitions Archive

  • Type: Competition Archive
  • Goal: Collection of solutions from SCU-hosted AI competitions
  • Tech: XGBoost, CatBoost, RandomForest, Sklearn
  • Highlights:
    • Baseline + advanced models
    • Analysis of model performance per task
  • Status: ✅ Completed
  • GitHub Repo

🚢 Titanic Survival Prediction

  • Type: Classification (ML Beginner Project)
  • Goal: Predict passenger survival on Titanic dataset
  • Tech: Sklearn, Logistic Regression, RandomForest
  • Highlights:
    • EDA + basic feature engineering
    • Simple model pipeline for classification task
  • Status: ✅ Completed
  • GitHub Repo

🧪 DACON: Electricity Usage Prediction

  • Type: Time Series Regression
  • Goal: Predict electricity consumption per building per hour
  • Tech: LightGBM, XGBoost, Optuna, Stacking
  • Highlights:
    • Feature engineering + voting ensemble
    • Hyperparameter tuning via RandomizedSearchCV
  • Status: 🛠 WIP
  • GitHub Repo

📂 More coming soon…

Stay tuned for updates on upcoming CV experiments and paper reproductions!