Highlights
What I built
- Pose landmark extraction across user and reference clips.
- Temporal alignment layer for movement comparison.
- Full-stack upload, processing, and feedback interface.
← Back to Projects
Machine LearningIn progress
Bhangra Coach
Computer vision dance coach with pose comparison feedback.
This project explores whether movement feedback can feel useful instead of robotic. The pipeline compares user dance clips against references and surfaces actionable differences.
- Date
- 2026
- Signal
- CV + ML
- Build stage
- Actively iterating on feedback quality
- Stack
- Next.js, FastAPI
computer-visionmlfullstackproduct-iteration
Demo
Project notes
Architecture
How the system works
- Frontend handles upload and review flow.
- FastAPI service computes landmark and delta metrics.
- Supabase stores clip metadata and processing outputs.
Challenges
What made it hard
- Movement alignment is hard when tempo differs between performers.
- Raw pose deltas needed interpretation to become useful coaching cues.
Lessons
What I learned
- Feedback UX is as important as model pipeline quality.
- Domain context (dance technique) matters for feature design.
Stack / materials
Next.jsFastAPIMediaPipeSupabaseComputer Vision
- Current focus is improving robustness across camera angles and lighting.
Media timeline
Build photos, clips, and process visuals. The goal is to show how the project evolved, not just the final screenshot.
Build snapshot