Skip to content

Sagarshresti18/webapp

Repository files navigation

Multimodal Sentiment Analysis Web App 🎯 Overview This web application performs sentiment analysis across three modalities:

  • 📝 Text: Analyzes sentiment, subjectivity, and polarity of user input.
  • 🖼️ Image: Predicts sentiment from facial expressions using a trained image model.
  • 🎙️ Audio: Detects emotion from voice recordings using MFCC features and a deep learning model. Features
  • Text Sentiment Analysis: Uses NLP techniques to classify sentiment and measure subjectivity/polarity.
  • Image Sentiment Analysis: Predicts emotion from facial images using a pre-trained model.
  • Audio Sentiment Analysis: Extracts MFCCs and uses a Conv2D model to classify emotions like happy, sad, angry, etc.
  • Style Transfer: Transforms input text into different writing styles using fine-tuned pipelines. Notes
  • The image sentiment model (img_sentiment.pkl) and audio model (SER_model.h5) must be present in the root directory.
  • Uploaded files are stored in the uploads/ folder.
  • Style transfer uses predefined pipelines for tone transformation. Audio Input Guidelines
  • Format: .wav
  • Duration: 2–5 seconds
  • Content: Clear emotional speech (e.g., happy, sad, angry)
  • Sampling Rate: 22050 Hz
  • Mono channel preferred 🛠️ Tech Stack 💻 Frontend
  • HTML5 – Structure and layout
  • CSS3 – Styling and responsive design
  • Jinja2 – Templating engine for dynamic content rendering 🧠 Backend
  • Python 3.12
  • Flask – Lightweight web framework for routing and server logic 🧪 Machine Learning & NLP
  • Keras – Deep learning framework for building and loading models
  • Scikit-learn – Image sentiment model and preprocessing
  • TextBlob – Text sentiment, polarity, and subjectivity analysis 🎙️ Audio Processing
  • Librosa – Audio feature extraction (MFCCs)
  • NumPy – Numerical operations and array manipulation 🖼️ Image Processing
  • OpenCV (optional) – Image preprocessing and face detection
  • Pillow – Image loading and manipulation 🔤 Style Transfer
  • Custom Pipelines – Rule-based or model-driven text transformation 📦 Utilities
  • Werkzeug – Secure file uploads
  • OS – File system operations.

About

Multimodal Sentiment Analyzer is a Flask-based web app that interprets human emotion from text, images, and audio using machine learning models and a responsive HTML/CSS frontend.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors