SentimentSphere

SentimentSphere is a multimodal real-time emotion recognition system that integrates three powerful models to understand and classify human emotions based on visual, textual, and speech inputs.

---

Overview

This project is developed as part of a DC credit initiative and includes:

1. Visual Emotion Recognition (CNN)

• Captures live webcam feed.
• Uses a Convolutional Neural Network to classify facial emotions in real time.
• Recognized emotions include: Happy, Sad, Angry, Neutral, Surprised, and more.Roundtable

2. Text-Based Emotion Recognition (NLP + LSTM)

• Accepts a sentence as input.
• Predicts the tone and emotion conveyed using NLP techniques and an LSTM model.
• Example input: "I'm feeling great today!" → Emotion: Happy

3. Speech-Based Emotion Recognition (LSTM)

• Records live audio through microphone.
• Extracts features like MFCCs and feeds them to an LSTM model to classify the emotion in real time.
• Real-time feedback during conversation or speech.

---

Tech Stack

• Frontend/Hosting: Streamlit
• Backend Models: Python (TensorFlow, Keras, OpenCV, NLTK, Librosa)
• Other Tools: NumPy, Pandas, Matplotlib

---

How It Works

Each model runs independently but is hosted together on a unified Streamlit interface. Due to deployment constraints, the models are currently hosted locally.

---

Try It Locally

Clone the repo:

git clone https://github.com/tashir0605/SentimentSphere
cd SentimentSphere