π§ Daily Study Log [2025-08-07]
Pushed forward on multiple tracks today β from idea generation to paper implementation, and competition experiments. Solid momentum across the board.
Proposed a computer visionβbased feedback system that tracks physical changes (e.g., back width, chest shape, shoulder gains) using pose estimation and visual analytics.
This idea explores how fitness progress can be made visually measurable, not just by weight or subjective feelings.
π View idea
Continued studying for the TOEIC exam.
Almost done with the full course β feeling confident and more prepared!
Tested the existing segmented Voting ensemble (Time Γ Building Type) with all engineered features removed.
Goal was to benchmark the performance using only raw features.
Used the best_model structure and introduced clustering-based features on top of engineered ones.
Maintained VotingRegressor for final prediction.
Studied how the DoWhy framework is used in practice by running actual implementation using the paperβs dataset.
This hands-on study helped bridge theory and application.
Advanced work on the bass_seeker audio analysis system.
Still in early stages, but the core modules are in place. Further tuning and feature additions planned.
π View project
Read multiple papers on pose-based action quality assessment in Taekwondo.
This was part of a broader literature review for novelty identification and future positioning.
π Example paper
π Visual Body Tracker: New CV-based body progress feedback idea
π TOEIC: Almost completed full preparation
π Competition: Tested raw-only features and clustering-enhanced ensemble
π DoWhy: Ran practical causal inference experiments
π Bass Seeker: Built early prototype for audio-based recommender
π Taekwondo Papers: Deep literature scan for novelty mapping