🧠 Daily Study Log [2025-07-17]
Covered a wide range of topics today — from test prep to project implementation to idea generation.
A productive and well-balanced day!

📊 Smoking Zone GIS Project — Legal Area Mapping

Focus: Building a GIS-based visualization of legal smoking zones in Seoul
Goal: Provide a clear and interactive public resource that highlights legally permitted smoking areas, not just “self-declared” ones

Highlights:

• Finalized dataset acquisition from 서울 열린데이터 광장
• Converted address data to lat/lon via geocoding
• Built an interactive map using Folium (district-level overlays + individual markers)

🧠 TOEIC Study — Listening Practice

What I Did:

• Focused on Part 3/4 listening drills (workplace conversations + announcements)
• Timed section practice (45 minutes) with error review

Insights:

• Getting more comfortable with fast dialogues
• Distractors in questions are still tricky — need more focused review
• Continue grammar review (especially verb tense traps)

💡 Idea Brainstorm — fact_rank

What I Did:

• Sketched out the idea of a community-curated fact-checking platform, focused on verifiability
• Introduced a scoring mechanism via RLHF-like upvoting for “facts” vs. “claims”
• Tentative project title: fact_rank

Insights:

• Core issue is moderation & incentive design
• MVP could be academic reference-only fact aggregation
• Possibly valuable for digital literacy education

📄 Paper Review — Action Recognition Using Pose Estimation

Started: Action Recognition Using Pose Estimation (arXiv:2109.09166)
Today’s Focus: High-level architecture and relevance to dance action classification
Repo: To be added after summary notebook completion

Reflection:

• Uses OpenPose-based 2D keypoints + CNN (2D & 3D) for action recognition
• Avoids using raw RGB input — fully pose-based, lightweight approach
• Architecture is modular and well-suited for dance genre classification tasks
• Paper structure and scope is very manageable for an undergrad-level replication

🔍 Research Club Planning — Feasibility Check

What I Did:

• Revisited our short paper idea feasibility for an undergrad team
• Compared existing literature vs. expected technical stack
• Evaluated team capability for data collection, modeling, and presentation

Insights:

• A simplified version (e.g., genre classification via BlazePose + CNN) is definitely feasible
• Dataset creation is the key bottleneck — but solvable via filming or YouTube scraping
• Paper writing likely viable within 1–2 months if scoped properly

✅ TL;DR

📍 Project: Smoking zones visualized at legal level — groundwork completed
📍 Study: Steady TOEIC progress — listening speed improving
📍 Paper: Pose-based action recognition is feasible and relevant for our dance project