AI Exercise Trainer 🏋️‍♂️

An AI-powered personal trainer that tracks and counts repetitions for exercises in real-time using MediaPipe for pose detection and a custom LSTM model for exercise classification.

---

Features ✨

• Real-time Pose Detection: Tracks 33 body landmarks.
• Exercise Classification: Recognizes bicep_curl, push_up, squat, shoulder_press.
• Repetition Counter: Counts reps automatically using joint angles.
• Performance Metrics: Displays FPS for monitoring.

---

Quick Start 🚀

Prerequisites

• Python 3.x
• Webcam

Installation & Running the Application

1. Clone the repository:

git clone https://github.com/DevPatel1106/Repo_github.git
cd Repo_github

2. Install dependencies and Run:

pip install -r requirements.txt
python main.py

How It Works ⚙️

• Capture Video: Webcam feed processed in a loop.
• Pose Detection: Extracts 33 (x, y) landmark coordinates.
• Sliding Window: Last 30 frames fed into the LSTM model.
• Exercise Prediction: Model predicts the exercise class.
• Repetition Counting: Detects “up” & “down” stages using joint angles.

---

Visual Example 🖼️

Real-Time Pose Detection

• Tracks 33 key body landmarks using MediaPipe.
• Displays predicted exercise class, rep count, and FPS.

Tip: For best accuracy, ensure proper lighting and full-body visibility in the webcam frame.

---

Project Structure 📁

|- main.py – Runs video capture & integrates modules.

|- ExerciseAiTrainer.py – Core logic for classification & rep counting.

|- PoseModule2.py – Pose detection & angle calculations.

|- AiTrainer_utils.py – Utilities (image resizing, FPS display).

|- ML.ipynb – Notebook for model training.

|- requirements.txt – Python dependencies.

|_ lstm_model.keras, scaler.joblib, label_encoder.joblib – Pre-trained model & encoders.

---