This project provides a unified Gradio interface that supports both:

• Multi-Domain Image Classification using pre-trained specialized models
• Zero-Shot Image Classification using Google's SigLIP and SigLIP2 models

It enables users to perform image classification across a wide variety of tasks and also test open-vocabulary classification via a simple web interface.

Supports 18+ domains, including:

• Age classification
• Gender classification
• Emotion detection
• Deepfake quality assessment
• Dog breed classification
• Waste classification
• Food classification (Indian/Western)
• Traffic density
• Leaf disease detection (rice)
• Alphabet sign language detection
• Gym workout pose classification
• Bird species classification
• Clip Art 126, Painting 126, Sketch 126
• MNIST and Fashion MNIST
• Multi-source 121 classification

Each model is integrated via its own module for efficient classification using domain-specific architectures.

Uses two open-vocabulary models:

• google/siglip-so400m-patch14-384
• google/siglip2-so400m-patch14-384

How it works:

• Accepts a user-uploaded image
• Takes a comma-separated list of labels
• Compares prediction probabilities using both SigLIP and SigLIP2 models

git clone https://github.com/PRITHIVSAKTHIUR/SigLIP2-MultiDomain-App.git
cd SigLIP2-MultiDomain-App

You can install required packages using pip:

pip install -r requirements.txt

Minimal requirements.txt might include:

torch
transformers
gradio
Pillow

python app.py

This will launch a Gradio interface in your default web browser.

1. Use the sidebar to choose your desired classification task (e.g., Age, Dog Breed, Deepfake).
2. Upload an image.
3. Click "Classify / Predict" to see the result.

1. Upload an image.
2. Enter labels separated by commas (e.g., "cat, dog, horse").
3. Click "Run" to compare SigLIP and SigLIP2 outputs.

├── app.py                           # Main Gradio app
├── gender_classification.py        # Domain model modules
├── emotion_classification.py
├── dog_breed.py
├── deepfake_quality.py
├── ...
├── sketch_126.py
└── requirements.txt

• Domain-specific models (e.g., CNNs, ViTs) hosted locally or via Hugging Face
• SigLIP / SigLIP2 models from Google (used for zero-shot inference)

This project is licensed under the MIT License. See the LICENSE file for details.