Name: FinalProject_PPT (1)(5)
Uploaded: 2026-04-25
Duration: 10 min 21 s
Description: Department of Computer Science and Engineering VIGNAN’S LARA INSTITUTE OF TECHNOLOGY AND SCIENCE Project Review on Virtual Gym Coach: Your Intelligent Workout Assistant.

Department of Computer Science and Engineering VIGNAN’S LARA INSTITUTE OF TECHNOLOGY AND SCIENCE Project Review on Virtual Gym Coach: Your Intelligent Workout Assistant. 

Scene 1

Department of Computer Science and Engineering
           VIGNAN’S LARA INSTITUTE OF TECHNOLOGY AND SCIENCE

Project Review 
on
 Virtual Gym Coach: Your Intelligent Workout Assistant

Under The Esteemed Guidance of Mrs. Sk. Nageena Jani M.Tech Assistant Professor CSE Department

Presented By  Batch No: A15
 B. Rishika               22FE1A0512 D. Dinesh               22FE1A0544 Ch. Ajay Kumar     22FE1A0531 B. Vinay                 21FE1A0513

Department of Computer Science & Engineering

CONTENTS. 24/04/2026. 2. Department of Computer Science & Engineering. 

Scene 2

Title  Abstract  Introduction  Literature Survey  Existing System  Proposed System  System Architecture  Proposed Methodology  Conclusion  References

24/05/2021. Department of Computer Science & Engineering. 

Scene 3

Virtual Gym Coach: Your Intelligent Workout Assistant

ABSTRACT. 24/04/2026. 4. Department of Computer Science & Engineering. 

Scene 4

The Virtual Gym Coach project solves the problem of poor posture during home workouts. It prevents users from getting poor results and injuries from unnatural movements. The site enhances exercise to be more interactive and fun. It implements TensorFlow.js with MoveNet to detect pose in real-time. A 3D trainer avatar created using Three.js provides visual feedback. The Web Speech API gives real-time verbal corrections. The system enables users to make posture adjustments instantly. Vite, React, and Tailwind CSS are utilized for an instant and responsive interface. Supabase handles workout data safely. Users can monitor posture accuracy effortlessly. Users can also see repetition and history of progress. A personal dashboard encourages consistency in users. The solution is easy and intuitive. It is safe and trustworthy for any fitness level. Overall, the project provides an immersive, secure, and personalized fitness experience.

INTRODUCTION. 24/04/2026. 5. Department of Computer Science & Engineering. 

Scene 5

Exercising at home has become very common today because it provides users with comfort, flexibility, and the freedom to work out anytime without the need for a gym. However, one major drawback of home workouts is the difficulty in maintaining the correct posture while performing different exercises. When posture is not correct, it reduces the effectiveness of the workout and, in some cases, may even result in injuries to the user. Another problem users face is the lack of proper tracking, as it becomes difficult to count repetitions or monitor progress in real time without assistance. Without proper guidance and feedback, users may lose motivation and eventually struggle to remain consistent with their fitness goals. To overcome these challenges, the Virtual Gym Coach project has been developed as a smart, web-based fitness platform.

INTRODUCTION. 24/04/2026. 6. Department of Computer Science & Engineering. 

Scene 6

The system uses TensorFlow.js with MoveNet for accurate and real-time pose detection, ensuring users can immediately identify and correct mistakes. A 3D trainer avatar designed with Three.js is included in the system to act as a visual coach, demonstrating the correct way to perform exercises. The Web Speech API is integrated to provide instant voice feedback, allowing users to hear corrections while they are working out in real time. The application is built with Vite, React, and Tailwind CSS, which ensures that the platform is fast, simple to use, and responsive across devices. Finally, all user data such as posture correctness, number of repetitions, and session history is securely stored and managed by Supabase, giving users a personal dashboard for tracking progress and staying motivated.

LITERATURE SURVEY. 24/04/2026. 7. Department of Computer Science & Engineering. 

Scene 7

S.No Paper Title Methodology Drawbacks Proposed Methodology to overcome those drawbacks 1 Virtual Fitness Trainer using Artificial Intelligence
 Uses Mediapipe for detecting body keypoints and providing posture correction. High computation cost, lag during real-time processing, limited voice interaction. Use MoveNet (lightweight) for real-time pose detection with low latency, and add voice feedback. 2 Real-time AI posture correction for powerlifting exercises using YOLOv5 and MediaPipe. Uses YOLOv5 and MediaPipe, Random Forest, Transformer(classification) and streamlit. Needs multiple cameras, DL underperformance large dataset (MPED) is used for better performance.

LITERATURE SURVEY. 24/04/2026. 8. Department of Computer Science & Engineering. 

Scene 8

S.No Paper Title Methodology Drawbacks Proposed Methodology to overcome those drawbacks 3 Advances for indoor fitness tracking, coaching, and motivation Review Paper surveying technologies for indoor fitness. Limitation for aerobic exercises, IoT devices. Provide real-time feedback for both aerobic and anaerobic exercises. 4 LogRF:  An approach to human pose estimation  Uses LogRF, a selection and classification framework. Limited to 7 exercises, Lack of real time Interface. Dataset expansion, hybrid integration of models, Real-time GUI.

LITERATURE SURVEY. 24/04/2026. 9. Department of Computer Science & Engineering. 

Scene 9

S.No Paper Title Methodology Drawbacks Proposed Methodology to overcome those drawbacks 5 Accuracy and adaptability improvement in aerobic training Proposed a self-attention-based 3D pose estimation and kinematic modelling framework. Customized dataset with 100 videos, 20 participants. We provide a large dataset named MPED and integrate self-attention. 6 Real-Time Posture Correction of Squat Exercise Uses Mediapipe and Bi-RGU model for classification. Uses Stereo type cameras to capture the pose, Small dataset(1332 squats) We use smartphone camera instead of stereo cam and large dataset.

LITERATURE SURVEY. 24/04/2026. 10. Department of Computer Science & Engineering. 

Scene 10

S.No Paper Title Methodology Drawbacks Proposed Methodology to overcome those drawbacks 7 AI trainer: Autoencoder based approach Autoencoder,  Bi-GRU classification Correction via skeletal stick models and stereo vision setup. We provide 3D model for correction. 8 Remote monitoring of physical rehabilitation of stroke patients. Uses smart wearables (bands, sensors) to track movements, and heart rate. Expensive, requires external hardware, limited to users owning devices. Provide a software-only solution using webcam and AI, reducing dependency on external devices.

LITERATURE SURVEY. 24/04/2026. 11. Department of Computer Science & Engineering. 

Scene 11

S.No Paper Title Methodology Drawbacks Proposed Methodology to overcome those drawbacks 9 Application of IoT and VR in college Physical Education Used sensors and IoT platforms to generate the 3D scenes. Complex setup requirements and IoT devices. Develop lightweight and scalable architechture.

EXISTING SYSTEM. 24/04/2026. 12. Department of Computer Science & Engineering. 

Scene 12

Uses MediaPipe for detecting skeletal landmarks in real-time. Employs OpenCV for video capture and frame-by-frame posture analysis. Relies on machine learning algorithms to suggest basic exercise plans. Tracks workout progress such as completed reps and calories burned. Focuses on posture correction by comparing the user’s pose with an ideal pose. Offers on-screen alerts for incorrect posture during workouts.

EXISTING SYSTEM. 24/04/2026. 13. Department of Computer Science & Engineering. 

Scene 13

Works as a virtual fitness trainer without the need for a human coach. Provides a static, pre-defined database for exercise instructions. Supports goal-based exercise recommendations according to user input. Features a basic UI to select exercises and view performance metrics. Designed mainly for home-based fitness without specialized gym equipment.

EXISTING SYSTEM. 24/04/2026. 14. Department of Computer Science & Engineering. 

Scene 14

user lands on
user selects pose
(excercise)
Model estimates user
P OSe
Model
Model Calculates
estimated On
Angles
Compare angles using
Cosine similarity
Display reps and
Calories
Try another Pose

Existing System Drawbacks. 24/04/2026. 15. Department of Computer Science & Engineering. 

Scene 15

Limited Dataset & Generalization: Uses a restricted dataset, causing pose detection errors for complex exercises or unusual angles. Lack of Advanced 3D Pose Estimation: The system relies on 2D pose estimation only, leading to inaccurate feedback for depth-based movements. No Voice Feedback: Users have to rely on visuals, which may interrupt the workout flow. Static Virtual Interaction: No 3D virtual avatar or immersive coaching experience.

PROPOSED SYSTEM. 16. Department of Computer Science & Engineering. 

Scene 16

Users create personal accounts by registering with basic details and fitness goals.  Webcam-based pose detection using MoveNet (TensorFlow.js) tracks body joints in real time.  Real-time voice feedback via the Web Speech API alerts users to incorrect posture.  A 3D virtual avatar coach (Three.js) demonstrates exercises for clear visual guidance.  The system automatically counts repetitions and tracks sets. Posture accuracy is monitored and recorded for each exercise.

PROPOSED SYSTEM. 24/04/2026. 17. Department of Computer Science & Engineering. 

Scene 17

All workout data is securely stored in Supabase for individual users. Users can view a personalized dashboard showing session history. Progress tracking provides insights into performance improvements over time.  Interactive features ensure a safe and engaging home workout experience.  No additional devices are required, making the system fully accessible at home.

SYSTEM ARCHITECTURE. 24/04/2026. 18. Department of Computer Science & Engineering. 

Scene 18

Webcam
/ Mobile
Camera
Applicatjon
Layer
Pose Estjmatjon Nework
Full con
Conv
React + Tailwind IJI + Vite
3D Virtual Trainer
Supabase
Featured Maps
Featured Maps
Featured Maps
Posture Analysis & Feedback
Module
Ideal Pose Comparison
Gesture & Rep Analysis
Wrong Posture Detection
Voice Feed
Cluster
Output Layer
Visualizabon
Audio
Performance
Metrics
Feedback

PROPOSED METHODOLOGY. 24/04/2026. 19. Department of Computer Science & Engineering. 

Scene 19

Technologies Used:  Pose Detection (MoveNet + TensorFlow.js): Real-time body joint trackingVoice  Feedback (Web Speech API): Instant posture correction alerts 3D Virtual Avatar Coach (Three.js): Visual exercise demonstrations Repetition & Set Tracking: Automatic counting of reps and sets Posture Accuracy Monitoring: Detects and records correctness of form Supabase Database: Secure storage of user workout data Personalized Dashboard: Session history and progress visualization

PROPOSED METHODOLOGY. 24/04/2026. 20. Department of Computer Science & Engineering. 

Scene 20

Advantages of Proposed System: Real-Time Feedback through voice commands. Independence on External Hardware making this accessible for all users. Advanced 3D Trainer to perform actions accurately. Posture Accuracy Monitoring: Detects and records correctness of form Personalized Dashboards to track user’s session history and progress.

PROPOSED METHODOLOGY. 24/04/2026. 21. Department of Computer Science & Engineering. 

Scene 21

CONCLUSION. 24/04/2026. 22. Department of Computer Science & Engineering. 

Scene 22

The suggested system effectively integrates AI-based pose detection, real-time vocal feedback, and a 3D virtual trainer to provide an enjoyable, interactive, and safe home workout experience. Through automated repetitions, set tracking, and posture accuracy assurance, it offers users accurate guidance on par with a personal coach. With secure data storage in Supabase and a customized progress tracking dashboard, users are able to track their fitness progress efficiently. Above all, the system needs no specialized hardware other than a webcam, so it is extremely accessible and convenient for home use.

REFERENCES. 24/04/2026. 23. Department of Computer Science & Engineering. 

Scene 23

Gupta, Lakshay, Shrey Gurbuxani, and Kapil Madan. "Virtual Fitness Trainer using Artificial Intelligence." Proceedings of the 2024 Sixteenth International Conference on Contemporary Computing. 2024.
 Ko, Yeong-Min, Aziz Nasridinov, and So-Hyun Park. "Real-time AI posture correction for powerlifting exercises using YOLOv5 and MediaPipe." IEEE Access (2024).
 Wang, Tianyi, et al. "Advances for indoor fitness tracking, coaching, and motivation: A review of existing technological advances." IEEE Systems, Man, and Cybernetics Magazine 7.1 (2021): 4-14.
 Raza, Ali, et al. "LogRF: An approach to human pose estimation using skeleton landmarks for physiotherapy fitness exercise correction." IEEE Access 11 (2023): 107930-107939.
 Qu, Hang, et al. "Accuracy and adaptability improvement in aerobic training: integration of self-attention mechanisms in 3d pose estimation and kinematic modeling." IEEE Access 12 (2024): 112470-112481.

REFERENCES. 24/04/2026. 24. Department of Computer Science & Engineering. 

Scene 24

Rao, Prashant, C. S. Asha, and P. Raghavendra Rao. "Real-Time Posture Correction of Squat Exercise: A Deep Learning Approach for Performance Analysis and Error Correction." IEEE Access (2025).
 Chariar, Mukundan, et al. "AI trainer: Autoencoder based approach for squat analysis and correction." IEEE Access 11 (2023): 107135-107149.
 Postolache, Octavian, et al. "Remote monitoring of physical rehabilitation of stroke patients using IoT and virtual reality." IEEE journal on selected areas in communications 39.2 (2020): 562-573.
 Ding, Yu, Yuhang Li, and Lei Cheng. "Application of Internet of Things and virtual reality technology in college physical education." Ieee Access 8 (2020): 96065-96074.

Results. 24/04/2026. 25. Department of Computer Science & Engineering. 

Scene 25

+ Virtual Gym Coa
Weekly progress
Welcome back, Dinesh! Keep crushing your goals
stay
Exercise Breakdown

Virtual Gym Coach
Your A1-powered fitness companion
Get Started
Sign in to your mvkout
Sign In
Sign up
Stan fitness journey with A1-powered coaching

Results. 24/04/2026. 26. Department of Computer Science & Engineering. 

Scene 26

Current Exercise
Available Exercises
to
3
Q' A1 Coach Controls
to s titrøs

Future Enhancements. 24/04/2026. 27. Department of Computer Science & Engineering. 

Scene 27

Future Enhancements: 
 Interactive AR Model to communicate with the users. Accuracy enhancements for keypoints detections. Improve Voice commands and response. Adaptive workout plans based on Progress, fatigue levels and injury risks. AI chatbot for real time fitness queries. Leaderboards, Challenges, and badges. Connect with IoT devices, smartwatches, and other fitness bands.

24/04/2026. 28. Department of Computer Science & Engineering. 

FinalProject_PPT (1)(5)

Scene 1 (0s)

Scene 2 (21s)

Scene 3 (33s)

Scene 4 (46s)

Scene 5 (1m 28s)

Scene 6 (2m 6s)

Scene 7 (2m 42s)

Scene 8 (3m 9s)

Scene 9 (3m 33s)

Scene 10 (3m 59s)

Scene 11 (4m 24s)

Scene 12 (4m 39s)

Scene 13 (5m 1s)

Scene 14 (5m 19s)

Scene 15 (5m 30s)

Scene 16 (5m 52s)

Scene 17 (6m 15s)

Scene 18 (6m 32s)

Scene 19 (6m 45s)

Scene 20 (7m 8s)

Scene 21 (7m 25s)

Scene 22 (7m 42s)

Scene 23 (8m 10s)

Scene 24 (8m 51s)

Scene 25 (9m 24s)

Scene 26 (9m 39s)

Scene 27 (9m 50s)

Scene 28 (10m 10s)