Skip to content

mountstorm/Kansei

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 

Repository files navigation

Kansei 感性

Computer vision coaching for football players who train without coaches, cameras, or data.

by Muzaffar Khaydarov · Mountstorm Labs


What is Kansei?

Kansei (感性) is a Japanese concept meaning perceptual intelligence — the ability to sense and refine movement through feel and feedback. That's exactly what this app gives players who've never had access to it.

Point your phone camera at training. Kansei analyzes:

  • Kick biomechanics — joint angles, follow-through arc, body lean at contact
  • Penalty accuracy — ball trajectory mapped against target zones
  • Positional load — sprint intensity, distance covered, work rate by role
  • Role-specific metrics — GK distribution, CB tackle positioning, MF press coverage, FWD finishing patterns

No expensive hardware. No camera operators. No enterprise contracts. Just your phone.


The problem

Tools like Catapult, STATSports, and Hudl exist. They cost $3,000–$15,000 per team per season.

There are 50M+ youth football players in Central Asia, Southeast Asia, and Sub-Saharan Africa who train every day without a single piece of performance data. No coach tells them their plant foot is wrong. No replay shows them why the penalty went left. They improve slowly, by feel, or they don't improve at all.

Kansei is built for them.


How it works

Phone camera → pose estimation → biomechanics analysis → role-specific feedback
      ↓
Optional wearable (BLE) → sprint/load metrics → combined dashboard

Core pipeline:

  1. Video captured on standard phone (iOS / Android)
  2. On-device pose estimation via MediaPipe — no cloud dependency
  3. Joint angle extraction and movement pattern classification
  4. Role-specific scoring model (GK / CB / MF / FWD)
  5. Feedback delivered in-app — text plus visual overlay

Why on-device? A player in Fergana, Uzbekistan or Dhaka, Bangladesh shouldn't need reliable internet to get feedback on their penalty kick. Everything runs locally.


Tech stack

Layer Technology
Mobile React Native
Pose estimation MediaPipe Pose (on-device)
CV pipeline Python, OpenCV
ML models PyTorch
Optional wearable BLE sensor integration
Backend (planned) FastAPI, PostgreSQL

Current status

Pre-prototype — active build starting now.

  • Concept validated through lived experience (grew up playing football in Uzbekistan)
  • Technical foundation: distributed AI inference research with Prof. Feng Wang, University of Mississippi
  • Proof of fast shipping: ValueStop — AI optimizer shipped in 24hrs at Emory Hacks 2025
  • Parallel work: NeuraBash — on-device LLM inference, 30% RAM reduction
  • MediaPipe pose estimation prototype
  • Kick biomechanics classifier v0
  • iOS/Android MVP
  • First 10 users — youth clubs in Uzbekistan

Roadmap

Phase 1 — Core CV (now)

  • Pose estimation pipeline on recorded video
  • Kick angle and follow-through analysis
  • Basic accuracy scoring for penalty kicks

Phase 2 — Mobile MVP

  • React Native app, iOS first
  • Real-time feedback overlay
  • Player profile and session history

Phase 3 — Wearable integration

  • BLE wearable for sprint and load metrics
  • Role-specific dashboards
  • Team and coach mode

Phase 4 — Market expansion

  • Uzbekistan pilot with youth clubs
  • Localization: Uzbek and Russian
  • B2B: regional football academies

Why this exists

I grew up in Uzbekistan. Football is everything there — every neighborhood, every school, every empty lot has a game running. But coaching infrastructure is nearly zero outside the top professional clubs. Kids train for years on pure instinct with no feedback loop.

I moved to the US at 18 and started studying CS. I'm now researching how to run AI inference on low-cost edge devices. At some point the two things connected: the technical problem I'm working on in the lab is exactly what would let me build the product my teammates back home needed.

That's Kansei.


About the founder

Muzaffar (Victor) Khaydarov@mountstorm

CS junior, University of Mississippi (Sally McDonnell Barksdale Honors College, 4.0 GPA)
Undergraduate researcher — distributed deep learning inference, Prof. Feng Wang
CS Teaching Assistant — 49 students
Incoming SWE Intern — C Spire (Summer 2026)
Originally from Tashkent, Uzbekistan


Contact

GitHub: github.com/mountstorm
LinkedIn: linkedin.com/in/muzaffar-
Email: mkhaydar@go.olemiss.edu


Pre-prototype under active development. Star the repo to follow progress.

About

Computer vision coaching for football players who can't afford the pros.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors