Emotion-Based Music Recommender

This project is an Emotion-Based Music Recommender system that uses facial and hand landmarks detected through a webcam to infer the user's emotion and recommend music based on that emotion.

Features

• Real-time Emotion Detection: Utilizes Mediapipe's holistic model to detect facial and hand landmarks in real-time.
• Emotion Classification: A pre-trained Keras model (model.h5) classifies the detected landmarks into a specific emotion.
• Music Recommendation: Based on the detected emotion, the application suggests music by searching on YouTube.

Installation

1. Clone the repository**:

   git clone https://github.com/Gayathri-Selvaganapathi/emotion_based_music_recommendation.git
   cd emotion_based_music_recommendation

2. Install the necessary dependencies:

Ensure you have Python 3.7+ installed, then run:

pip install -r requirements.txt

3. Prepare the model and label files:

   • Place your pre-trained Keras model (model.h5) in the project directory.
   • Place your label file (labels.npy) in the project directory. This file should contain the class names corresponding to the emotions your model can detect.

Usage

1. Run the Streamlit application:

streamlit run app.py

2. Interact with the application:

   • The application will start the webcam and detect your emotion in real-time.
   • Once the emotion is detected, you can click the "Recommend me songs" button to open YouTube with a search query based on your emotion.
   • The application will search YouTube for songs that match your detected emotion.

Files

* app.py: The main script to run the application.
* model.h5: The pre-trained Keras model for emotion detection.
* labels.npy: The labels corresponding to the emotions.

How It Works

1. Emotion Detection: The application uses Mediapipe to process video frames from the webcam. It detects facial landmarks and hand landmarks.

2. Feature Extraction: The detected landmarks are used as features for emotion classification. These features are normalized relative to specific key points on the face and hands.

3. Prediction: The features are passed through a pre-trained Keras model that predicts the user's emotion.

4. Music Recommendation: Based on the predicted emotion, the application suggests music by generating a YouTube search query.