Student Innovation-Ideas that can boost fitness activities and assist in keeping fit.

Solution Dossier & Technical Blueprint: AICTE Smart Fitness & Activity Platform

1. Executive Snapshot

Summary

The proposed solution, FitPulse AI, is an edge-AI powered, privacy-preserving physical fitness and campus engagement platform designed to boost daily physical activity among students and citizens. The system addresses low physical activity adoption in higher education institutions (HEIs) by combining real-time, on-device Computer Vision pose evaluation (via WebGL/WASM) for automated repetition counting and posture correction with an institutional gamification engine natively aligned with the AICTE Activity Point Program [RESEARCHED] and the Fit India Movement Age-Appropriate Fitness Protocols (GOALS 18–65) [RESEARCHED]. Operating as a progressive web application (PWA), the system requires zero expensive wearable hardware or high-speed internet, executing frame analysis locally on standard smartphones and laptop cameras while maintaining complete user video privacy [INFERENCE].

Metadata Table

Fundamental System Objective

To deliver a zero-hardware-cost, privacy-first computer vision and gamification platform that turns any smartphone into a real-time fitness coach while automating fitness activity verification for institutional AICTE Activity Points.

2. Problem Deconstruction

The explicit problem description provided in the SIH portal is:

"Student Innovation-Ideas that can boost fitness activities and assist in keeping fit." [PS]

Atomic Requirements & Implementation Implications

• R1: Focus on Student & Youth Innovation [PS]

  • Interpretation: The solution must target student lifestyle patterns, address academic sedentary behavior, and fit within institutional frameworks of Indian colleges and universities.
  • Implementation Implication: Embed native workflows for earning AICTE Activity Points (where 100 points are mandatory for 4-year B.Tech degree completion) [RESEARCHED] through physical fitness and inter-collegiate challenges.
  • Confidence Tag: [INFERENCE]

• R2: Boost Physical Fitness Activities [PS]

  • Interpretation: The platform must provide active motivation mechanisms (gamification, social proof, inter-departmental leagues, daily streaks, rewards) to convert passive users into active participants.
  • Implementation Implication: Implement competitive leaderboards, group fitness challenges ("Sundays on Cycle", "Daily 30-Min Workout"), and campus fitness feeds.
  • Confidence Tag: [INFERENCE]

• R3: Assist in Keeping Fit (Guidance & Feedback) [PS]

  • Interpretation: The system must act as an automated guide, helping users perform exercises correctly, preventing injury through real-time feedback, and providing structured fitness routines tailored to individual capability.
  • Implementation Implication: Integrate real-time human pose estimation (using 33 3D keypoint tracking) [RESEARCHED] for posture validation, live rep counting, and joint angle evaluation.
  • Confidence Tag: [INFERENCE]

• R4: Software-Only Category Constraint [PS]

  • Interpretation: The solution must not depend on custom manufactured hardware, proprietary smartwatches, or expensive sensor setups.
  • Implementation Implication: Leverage standard RGB camera sensors on existing Android/iOS devices and webcams via WebGL and WebAssembly (WASM) for client-side processing.
  • Confidence Tag: [INFERENCE]

3. Stakeholders & User Needs

4. Requirements Traceability

5. Constraints & Environment Realities

• Geography & Field Conditions: Students and citizens in Indian colleges often work out outdoors, in hostels with limited indoor space, or on college grounds under varying lighting conditions (bright sunlight, low-light hostel rooms) [KNOWN].
• Hardware Diversity: Target users possess smartphones ranging from budget Android devices (2 GB–4 GB RAM, entry-level MediaTek/Snapdragon processors) to high-end laptops [KNOWN]. Deep neural networks requiring heavy GPU computation will fail on budget devices.
• Connectivity & Bandwidth: Campus Wi-Fi and mobile networks in tier-2/tier-3 cities frequently experience packet loss and bandwidth throttling [KNOWN]. Streaming HD camera video to cloud servers for computer vision inference is infeasible and cost-prohibitive.
• Privacy Concerns: Users, especially female students, are unwilling to record and upload video clips of themselves working out to third-party cloud servers [KNOWN]. Video frames MUST be processed strictly in volatile device memory and discarded immediately.
• Language & Literacy: Users across different regions require multi-lingual audio-visual guidance (English, Hindi, regional languages) with simple visual skeletal overlays rather than text-heavy instructions [INFERENCE].

6. Research Questions

1. How fast can state-of-the-art light-weight pose estimation models run on budget mobile browsers?
   • Result: MediaPipe Pose (BlazePose Lite) achieves ~20–30 FPS on mid-range Android devices via WebGL/WASM, and MoveNet Lightning achieves ~30 FPS on CPU/WASM [RESEARCHED].
2. What are the official fitness protocols mandated by the Indian government for the 18–65 age group?
   • Result: The Fit India "Age-Appropriate Fitness Protocols (GOALS 18–65)" focus on Flexibility (Sit & Reach), Core Strength (Partial Curl-Up), Muscular Endurance (Push-Ups for Men, Modified Push-Ups for Women), and Body Composition (BMI) [RESEARCHED].
3. How are AICTE Activity Points allocated for sports and fitness?
   • Result: Under the AICTE Activity Point Program, students require 100 points (75 for lateral entry). Participating in sports and fitness initiatives yields 10–20 points per verified category activity (requiring ~80-90 hours or structured multi-week logging) [RESEARCHED].

7. Data Requirements & Sources

8. Existing Solutions & Gap Analysis

9. Candidate Solution Approaches

Approach 1: Wearable & Sensor-Based IoT Tracking

• Mechanism: Use smart wristbands (IMU sensors) connected via Bluetooth Low Energy (BLE) to a mobile app to detect movement patterns.
• Problem Fit: Good for heart rate and step tracking; poor for multi-joint pose assessment (squat depth, elbow flare).
• Complexity: High (requires hardware procurement, BLE firmware debugging, device pairing issues).
• Cost: High (₹1,500–₹3,000 per student; infeasible for mass university deployment).
• Hackathon Feasibility: Very Low (hardware dependency risks live demo failure).

Approach 2: Server-Side Cloud Video AI Analysis

• Mechanism: User records workout video on phone, uploads file to AWS/GCP, and a server-side PyTorch model processes the video using YOLOv8-Pose and returns form scores.
• Problem Fit: High accuracy model potential (can use larger deep learning architectures).
• Complexity: Medium-High.
• Cost: High (Cloud GPU costs for video transcoding and inference: ~$0.05 per video upload).
• Privacy & Network Risk: Extreme (uploading large MP4 files fails on spotty 3G/4G campus networks; user video uploaded to cloud violates student privacy).
• Hackathon Feasibility: Medium.

Approach 3: Edge-AI Computer Vision + Progressive Web App (PWA) + AICTE Gamification Framework (RECOMMENDED)

• Mechanism: A multi-platform Progressive Web Application (PWA) running Google MediaPipe Pose directly in the browser via WebGL/WASM [RESEARCHED]. Joint angle trigonometry and biomechanical state machines process pose keypoints locally on the device at 30 FPS. Only lightweight JSON activity telemetry (reps, duration, posture score) is synced to a lightweight backend.
• Problem Fit: Perfect (zero hardware cost, instant access via URL/QR code, total video privacy, low bandwidth).
• Complexity: Medium (requires robust client-side biomechanical state machine logic).
• Cost: Extremely Low (runs on free tier infrastructure; client device handles computation).
• Hackathon Feasibility: High (100% buildable and live-demonstrable using standard laptops/smartphones).

10. Recommended Approach & Why

Selected Approach: Approach 3 (Edge-AI PWA + Local Biomechanical State Machine + AICTE Activity Point Sync)

Why Alternatives Were Rejected:

• Approach 1 (IoT/Wearables) was rejected because requiring students to purchase smart wearables violates the principle of keeping fitness accessible and free ("Fitness as easy, fun, and free" - Fit India Mission) [RESEARCHED].
• Approach 2 (Cloud Video Processing) was rejected due to heavy cloud GPU hosting expenses, network latency in hostels, and severe student privacy resistance regarding video clip uploads.

Winning Edge of Approach 3:

1. Zero Setup Cost: Works instantly on any camera-enabled smartphone or laptop browser via WebGL.
2. Privacy by Design: Camera frames never leave RAM memory; only numerical landmark vectors (X, Y, Z coordinates) are evaluated locally.
3. Institutional Alignment: Seamlessly logs validated fitness hours directly into the college's AICTE Activity Point approval portal.

11. Technical Architecture

Layered Architecture Diagram

+-----------------------------------------------------------------------------------+
|                               USER INTERFACE LAYER                                |
|  React 18 PWA + Tailwind CSS + Canvas Overlay (Skeleton render at 60 FPS)         |
+-----------------------------------------------------------------------------------+
                                         |
                                         v
+-----------------------------------------------------------------------------------+
|                             ON-DEVICE VISION ENGINE                               |
|  MediaPipe Pose (WASM / WebGL) -> Extracts 33 3D Landmark Vectors @ 30 FPS         |
|  Biomechanical State Machine Engine -> Joint Angle Trigonometry & Rep Counter     |
+-----------------------------------------------------------------------------------+
                                         |
                                         v
+-----------------------------------------------------------------------------------+
|                        LOCAL STORAGE & OFFLINE ENGINE                             |
|  IndexedDB (Local Activity Logs & Draft Certificates) + Workbox Service Worker     |
+-----------------------------------------------------------------------------------+
                                         |
                       (Background Sync when Online)
                                         v
+-----------------------------------------------------------------------------------+
|                             BACKEND API GATEWAY                                   |
|  Node.js / Express REST API + JWT Auth + Rate Limiter                             |
+-----------------------------------------------------------------------------------+
        |                                   |                                  |
        v                                   v                                  v
+---------------+                   +---------------+                  +---------------+
|  POSTGRESQL   |                   |  REDIS CACHE  |                  | CRYPTO SIGNER |
| Relational DB |                   | Leaderboards  |                  | Activity Hash |
| Users & Logs  |                   | & Session Store|                 | Verification  |
+---------------+                   +---------------+                  +---------------+

Data Flow Narrative

1. Camera Stream Capture: The user opens the PWA and selects an exercise (e.g., Squats, Push-Ups, Planks, or Fit India Assessment Test). The app requests camera access via navigator.mediaDevices.getUserMedia().
2. Local Frame Processing: The HTML5 Video Element feeds raw frames directly into the in-browser MediaPipe Pose WASM runtime.
3. Landmark Extraction: MediaPipe predicts 33 keypoint 3D spatial coordinates $(x, y, z, \text{visibility})$ locally in $<15\text{ms}$ per frame [RESEARCHED].
4. Biomechanical Evaluation:
   • The app's math module calculates vector angles (e.g., Hip-Knee-Ankle angle for Squats, Shoulder-Elbow-Wrist angle for Push-ups).
   • A finite state machine (FSM) transitions states: START $\rightarrow$ DESCENDING $\rightarrow$ BOTTOM_THRESHOLD (Depth Check) $\rightarrow$ ASCENDING $\rightarrow$ REP_COMPLETE.
   • Form checks (e.g., "Knees caving in", "Back sagging", "Incomplete depth") trigger real-time audio and visual visual overlays on the HTML5 Canvas.
5. Session Aggregation & Cryptographic Signing: On workout completion, the client generates a session summary payload containing rep count, time duration, average posture score, timestamp, device telemetry, and optional coarse GPS coordinates. The payload is signed with a SHA-256 client session token.
6. Offline Queue or Cloud Sync:
   • If offline: The payload is saved to IndexedDB. A Service Worker registers a background-sync task.
   • If online: The client posts JSON payload to /api/v1/workouts/submit.
7. Backend Validation & Gamification: The Node.js API verifies token integrity, logs the entry into PostgreSQL, updates the user's weekly AICTE Activity Point hours, and updates live campus leaderboards stored in Redis.

12. Technology Stack

13. AI/ML Justification

Component 1: On-Device Pose Landmark Detection (MediaPipe Pose)

1. Exact Problem AI Solves: Converts raw 2D pixel matrices from camera frames into standardized 3D keypoint coordinates $(x, y, z)$ of human joints [RESEARCHED].
2. Input: 640x480 RGB Video Frames from camera stream.
3. Output: Array of 33 Keypoints with 3D coordinates and confidence visibility scores.
4. Why Simpler Rules Fail: Pixel-based color thresholding or traditional edge detection cannot robustly identify human joints under varying clothing, backgrounds, and lighting. Neural network regression is strictly required.
5. Data Required: Pre-trained BlazePose weight checkpoint (~6 MB).
6. Evaluation Metric: Mean Per Joint Position Error (MPJPE) and Percentage of Correct Keypoints ([email protected]).
7. Failure Mode Handling: If keypoint visibility falls below 0.6 (e.g., body partially out of frame), the UI displays "Step back - full body must be visible".
8. Training/Fine-tuning: None needed during hackathon; use pre-trained Google MediaPipe BlazePose Lite model [RESEARCHED].
9. Human Oversight: User can manually pause or terminate tracking if misclassifications occur.

Where AI Should NOT Be Used:

• Repetition Counting & Depth Validation: DO NOT use a black-box deep learning model to count reps or judge depth. Use deterministic trigonometric math (joint angles calculated via vector dot products) on top of AI keypoints. This guarantees 100% explainable, predictable, and debuggable rep counting logic.
• Activity Point Awarding: Rule-based verification engines must handle credit calculation based on verified duration and rep standards.

14. Data Architecture & Offline/Low-Network Design

Core Database Schema (PostgreSQL)

-- Users Table
CREATE TABLE users (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    full_name VARCHAR(100) NOT NULL,
    email VARCHAR(100) UNIQUE NOT NULL,
    role VARCHAR(20) CHECK (role IN ('student', 'faculty', 'admin', 'citizen')),
    college_id VARCHAR(50),
    department VARCHAR(50),
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

-- Fitness Activities Log Table
CREATE TABLE activity_logs (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id),
    exercise_type VARCHAR(50) NOT NULL, -- e.g. 'squat', 'pushup', 'fit_india_test'
    rep_count INT NOT NULL,
    duration_seconds INT NOT NULL,
    accuracy_score NUMERIC(5,2), -- Percentage correct posture
    telemetry_hash VARCHAR(64) NOT NULL, -- Cryptographic proof signature
    synced_from_offline BOOLEAN DEFAULT FALSE,
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

-- AICTE Activity Points Ledger Table
CREATE TABLE aicte_activity_ledger (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID REFERENCES users(id),
    points_awarded INT NOT NULL,
    category VARCHAR(100) DEFAULT 'Sports and Fitness Activities',
    verified_by_faculty UUID REFERENCES users(id),
    status VARCHAR(20) DEFAULT 'pending' CHECK (status IN ('pending', 'approved', 'rejected')),
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

Local Storage & Offline Synchronization Flow

• Local Caching: Application Shell (HTML/JS/CSS) and WASM binaries (~6MB) are cached locally via Service Worker Cache API during initial load.
• Offline Queueing: Workout sessions completed while offline are written directly to IndexedDB under object store offline_workouts.
• Sync Strategy: Upon detecting navigator.onLine == true, the PWA triggers a sync pipeline:
  1. Fetch un-synced records from IndexedDB.
  2. Send batch request to /api/v1/sync/batch.
  3. Server validates timestamps and telemetry hashes.
  4. On successful 200 OK response, local IndexedDB items are marked as synced: true.
• Conflict Handling: Server uses client-generated immutable UUIDs. If a record already exists in PostgreSQL, duplicate submissions are ignored cleanly (idempotent write).

15. Security, Privacy & Governance

• Camera Feed Privacy: Camera frames are strictly processed in device volatile RAM memory via HTML5 Canvas context. No image or video data is ever saved to disk or transmitted across the network.
• DPDP Act 2023 Compliance: Complies with India's Digital Personal Data Protection Act 2023 [KNOWN]. Explicit consent prompts notify users that health telemetry (rep count, activity duration) is collected solely for fitness scoring and activity point credit.
• Anti-Spoofing & Fraud Prevention:
  • To prevent users from shaking phone or playing pre-recorded videos in front of camera:
    • Temporal Kinematic Analysis: The system checks natural velocity and acceleration profiles of joint angles. Instantaneous static jumps indicate pre-recorded loop injection.
    • Challenge Mode: Prompts user to perform a randomized prompt during workout (e.g., "Raise left hand now") to prove live participation during institutional tests.
• Regulatory Disclaimer: The system is an educational and fitness engagement platform. Fitness assessments based on Fit India Guidelines are informational and do not constitute certified medical advice [INFERENCE].

16. MVP Scope (Hackathon Build)

In Scope

• Real-time AI Workout Coach: Client-side computer vision tracking for 3 core exercises: Squats, Push-Ups, and Planks.
• Live Form Guidance: Visual skeleton overlay on camera video + real-time posture error detection (e.g., "Keep hips lower").
• Fit India Protocol Test Module: Automated 1-minute Push-Up & Squat assessment based on GOALS 18–65 protocols [RESEARCHED].
• AICTE Activity Point Dashboard: Student view showing earned hours/points + Faculty approval interface mock.
• Campus Leaderboard: Live ranking by college/department using Redis sorted sets.
• Offline PWA Capability: Full workout functionality offline with automatic background sync upon re-connecting.

Described Only / Out of Scope

• Integration with proprietary smartwatch APIs (Apple HealthKit / Google Health Connect sync).
• Automated faculty signature certificate PDF generator with institutional digital signatures.
• Advanced 3D avatar coaching rendering in WebGL.

17. Live Demo Simulation Plan

To ensure a 100% flawless presentation during SIH judging, the demo plan will accommodate potential venue constraints (poor internet, bad lighting, lack of space):

1. Simulated WebCam Input (Fallback): Pre-record a sample 1080p MP4 video of a student performing correct and incorrect squats. Add a "Toggle Mock Video Stream" button in the PWA that routes the video file into the MediaPipe Pose pipeline instead of live webcam. This guarantees a live computer vision demo even if demo room lighting or space is inadequate.
2. Offline Mode Demonstration:
   • Step 1: Open app online, perform a 5-rep squat workout.
   • Step 2: Disconnect laptop/phone Wi-Fi live in front of judges.
   • Step 3: Perform another exercise; show local IndexedDB storage update.
   • Step 4: Re-enable Wi-Fi; show real-time background sync and campus leaderboard rank jump.
3. Faculty Audit Flow: Switch role to "Faculty Supervisor" and show real-time incoming workout proof logs pending approval for AICTE Activity Points.

18. Implementation Roadmap

Target Completion: 30 September 2026 [PS]

Phase 1: Architecture & Pose Engine (Weeks 1-2)
  ├── Set up Vite + React + Tailwind PWA shell
  ├── Integrate MediaPipe Pose WASM runtime
  └── Implement joint angle calculation library (trigonometric vector math)

Phase 2: Exercise State Machines & Core UI (Weeks 3-4)
  ├── Build finite state machines for Squat, Pushup, Plank
  ├── Develop HUD overlay canvas (skeleton rendering & rep counter)
  └── Implement Fit India Protocol test routine

Phase 3: Backend API, Auth & Offline Engine (Weeks 5-6)
  ├── Develop Node.js REST API + PostgreSQL schema
  ├── Build IndexedDB local storage + Workbox Service Worker sync
  └── Integrate Redis leaderboards

Phase 4: AICTE Portal & Testing (Weeks 7-8)
  ├── Build Student & Faculty Activity Point workflows
  ├── Conduct anti-spoofing tests & pose accuracy benchmark
  └── Final demo simulation prep & freeze (Before 30 Sept 2026)

19. Testing & Validation Plan

• Functional Verification: Test rep counting state machine across 100 repetitions of varied speeds (fast, slow, paused) to ensure zero missed or double counts.
• Pose Accuracy Testing: Benchmark MediaPipe keypoints against Kaggle Workout Pose Dataset [RESEARCHED] to verify joint angle calculation error is within $\pm 3^\circ$.
• Low-Light & Angle Robustness: Test camera tracking across 5 camera placement angles ($0^\circ$ frontal, $45^\circ$ diagonal, $90^\circ$ side-profile) and low-light hostel room environments ($<100\text{ lux}$).
• False Positive / Negative Testing:
  • False Positive Test: Have user jump or wave hands; verify squat counter does NOT increment.
  • False Negative Test: Have user do deep squats with partial side occlusion; verify rep IS correctly registered if depth threshold is crossed.
• Offline Resilience: Disconnect network mid-workout, complete session, force app crash/reload, verify data persists in IndexedDB and syncs upon network restoration.

20. Risk Register

21. Cost & Scalability

Hackathon

• Hosting: Free Tier (Vercel / Render for Node.js API).
• Database: Free Tier managed PostgreSQL (Supabase / Neon DB).
• Hardware Cost: ₹0 (Uses team's existing smartphones and laptops).
• Total Cost: ₹0.

Production (Scaling to 100,000 Students across 500 Colleges)

• Compute / API Servers: $2 \times$ AWS EC2 t4g.medium instances (Load balanced Node.js API): ~$30/month [INFERENCE].
• Database: AWS RDS PostgreSQL (db.t4g.small multi-AZ): ~$50/month [INFERENCE].
• Cache: AWS ElastiCache Redis (cache.t4g.micro): ~$15/month [INFERENCE].
• CDN & Static Asset Storage: Cloudflare Free / AWS S3 + CloudFront: ~$10/month [INFERENCE].
• Client Device Costs: ₹0 (All AI computation happens on student phones).
• Estimated Monthly Infrastructure Cost: $105/month (₹8,700/month) for 100,000 active users [INFERENCE].

22. Differentiation

23. Future Scope

1. Multi-Person Camera Tracking: Extend pose state machine to track multi-student physical education classes or inter-hostel group competitions on a single laptop camera.
2. Wearable Integration via Web Bluetooth API: Optionally allow users with smart heart-rate monitors to pair directly to the browser app for cardiovascular intensity logging.
3. National AICTE Analytics Integration: Provide a macro-level dashboard for MIC/AICTE to view national campus physical activity indices and health trends.

24. Assumptions & Open Items Log

Consolidated Assumptions & Items Needing Validation

• [ASSUMPTION] College faculty supervisors will accept digital cryptographic activity logs generated by the system for AICTE Activity Point approval.
• [ASSUMPTION] Student devices will have at least one working camera capable of 640x480 resolution video capture at 20+ FPS.
• [NEEDS VALIDATION] Direct REST API integration with official government Fit India portals or university LMS portals (e.g., SAMARTH ERP) requires official clearance and MOU agreements between AICTE and the institute [INFERENCE].

Final Recommendation

Build FitPulse AI as an Edge-AI Progressive Web App (PWA) using client-side MediaPipe Pose keypoint tracking, deterministic biomechanical state machines, and an AICTE Activity Point management module. This approach completely avoids hardware procurement costs, protects student video privacy, operates seamlessly offline, and directly solves the motivation and verification bottlenecks in higher education institutions.

Recommended Stack

• Frontend: React 18, Vite, Tailwind CSS, Workbox PWA.
• Vision Engine: MediaPipe Pose (WASM / WebGL runtime).
• Backend: Node.js, Express REST API.
• Database & Cache: PostgreSQL, Redis.
• Offline Storage: IndexedDB (idb wrapper).

Biggest Dependency

User device browser support for WebGL and MediaDevices Camera API (supported in $>98%$ of modern mobile browsers [KNOWN]).

Biggest Risk

Poor phone placement or backlighting in student hostel rooms causing degraded keypoint confidence. (Mitigated via on-screen positioning guides).

Biggest Technical Challenge

Implementing deterministic, lag-free biomechanical angle calculations and finite state machines in JavaScript running at 30 FPS alongside HTML5 Canvas rendering.

First 5 Actions

1. Initialize Vite + React PWA repository and configure Workbox Service Worker.
2. Import MediaPipe Pose WASM npm package and establish live webcam landmark stream.
3. Write vector trigonometry helper functions to compute 3D joint angles (Hip-Knee-Ankle, Shoulder-Elbow-Wrist).
4. Implement the Squat finite state machine (FSM) with rep counter and depth validation logic.
5. Build local IndexedDB schema for offline session caching and background sync API.

Validation Before Coding

• Validate that MediaPipe Pose Lite runtime achieves $\ge 20$ FPS on target test devices (budget Android smartphone and mid-range laptop).
• Verify joint angle threshold parameters for deep squats ($<90^\circ$ knee angle) and push-ups ($<90^\circ$ elbow angle) using sample exercise footage.

Buildability Score /100

95/100 (Extremely high hackathon buildability; 100% software-based, zero hardware dependencies, fully demonstrable on a single laptop/phone).

Confidence

HIGH (Supported by official AICTE Activity Point guidelines, Fit India GOALS protocols, and proven WebGL computer vision performance benchmarks).

Research Sources

1. Fit India Movement Official Guidelines & Protocols

   • URL: https://yas.gov.in/en/fit-india-fitness-protocols [RESEARCHED]
   • Source Type: Official Government Portal (Ministry of Youth Affairs and Sports)
   • Claim Supported: Verifies official Age-Appropriate Fitness Protocols (GOALS 18–65 years) for push-ups, squats, and flexibility testing.
   • Reliability: 100% (Primary Government Source).

2. Fit India Mobile App Details

   • URL: https://play.google.com/store/apps/details?id=com.sai.fitIndia [RESEARCHED]
   • Source Type: Official App Store Directory (Sports Authority of India)
   • Claim Supported: Documents existing feature set and gaps in current government fitness mobile app.
   • Reliability: 100% (Official App Distribution Channel).

3. AICTE Activity Point Program Guidelines

   • URL: https://vtu.ac.in/pdf/regulations2021/anex7.pdf [RESEARCHED]
   • Source Type: Institutional Circular & Regulatory Guidelines
   • Claim Supported: Validates requirement of 100 Activity Points for 4-year B.Tech students (75 for lateral entry) including sports and fitness activities.
   • Reliability: 100% (Official Academic Regulatory Document).

4. MediaPipe BlazePose On-Device Performance Analysis

   • URL: https://research.google/blog/on-device-real-time-body-pose-tracking-with-mediapipe-blazepose/ [RESEARCHED]
   • Source Type: Technical Research Publication (Google Research)
   • Claim Supported: Confirms real-time 33 3D keypoint tracking performance via WebGL/WASM on mobile web browsers at $\ge 30$ FPS.
   • Reliability: 100% (Official Developer & Research Documentation).