"Tarak" (तारक) in Sanskrit means Savior or Protector, and AI represents the Artificial Intelligence core of this project. tarak.AI is an interactive web application designed to tackle the problem of unreported river pollution by empowering the public to participate in environmental monitoring.

---

Project Overview

tarak.AI allows users to report river pollution in real time, primarily by uploading images. These images are processed by the InstructBLIP Vision-Language Model, which helps in:

• Determining the severity of pollution
• Identifying and classifying visible pollutants

Additional features include( some to be Integrated in future):

• Manual or automated (via EXIF data) location tagging
• Generation of pollution reports for public research
• Local and searchable air/water pollution statistics (Google Maps API integration)
• Government initiatives displayed in a dynamic slideshow format
• Multi-language support (via Bhashini integration) for inclusivity across Indian communities

---

Current Challenges

• Deployment: Handling model dependencies and real-time performance
• Image Processing Workflow: InstructBLIP is computationally intensive
• Model Integration:
  • Beginner approach: Use pre-trained InstructBLIP
  • Learning goal: Fine-tune the model using datasets like TACO, with guidance from the SLCR team

---

Future Enhancements

• Creation of a mobile app with all core features and a social media-like interface
• Integration of government APIs for real-time data and program insights
• Fine Tuning the model
• Backend Database Integration
• Gamification/Rewarding System

---

GitHub Repository

🔗 tarak.AI GitHub Repo 🔗 [Live Demo]- https://atharwaaah.github.io/tarak.AI

---

---

Setup Instructions

This project currently runs through a Google Colab notebook. Follow the steps below to get started:

1. Open the Colab Notebook

• Visit the notebook: 🔗 tarak.AI Colab

2. Run All Cells

• Ensure GPU runtime is enabled:
  Go to Runtime > Change runtime type > Select GPU
• Click Runtime > Run all to execute the notebook

3. Upload Images

• Use the provided upload cell to submit river pollution images
• The notebook will process them using the InstructBLIP Vision-Language Model and provide output on severity and pollutant type

---

Requirements (For Local Setup)

If running locally (optional), ensure the following:

• Python 3.8+
• pip
• Jupyter or VS Code
• Required libraries (install via pip install -r requirements.txt)

---

Author

Atharva Sachin Gupta
22024006, 3rd Year - IDD Biomedical Engineering
✉️ atharva.gupta.bme22@iitbhu.ac.in