INCOIS Ocean Hazard Reporting Platform

Project Overview

The INCOIS Ocean Hazard Reporting Platform is a specialized solution developed for Smart India Hackathon 2025 (Problem Statement: SIH25039) to facilitate the monitoring and reporting of ocean-related hazards. The platform provides a streamlined interface for field reporting and a robust administrative backend for verification and data analysis.

Core Pipeline and Verification

This project implements a hazard reporting and verification pipeline. Users submit incident reports through the mobile application, providing a hazard type, textual description and an image uploaded via Cloudinary. The report is persisted in MongoDB with an initial status of pending and an aiConfidenceScore of zero, and the client receives an immediate acknowledgement.

An asynchronous AI verification pipeline (server/services/aiVerification.js) then processes each report in two stages:

• CLIP-Based Image-to-Text Verification: The backend calls Hugging Face's inference API using the openai/clip-vit-large-patch14 model in zero-shot image classification mode. The submitted image is evaluated against a set of ocean-hazard candidate labels (e.g., ocean, waves, surge, flood). If the top-scoring label exceeds a similarity threshold of 0.70, up to 70 additional confidence points are awarded proportionally to the model's score.

These checks yield a composite aiConfidenceScore out of 100. Reports scoring 85 or above are automatically promoted to verified status. Reports below this threshold remain pending for manual review by an administrator or analyst, who can update the status via the PATCH /api/reports/:id/status endpoint. This multi-stage scoring approach—combining semantic image-text matching, reduces false positives and ensures only credible ocean hazard reports advance through the system.

Key Features

• Mobile Application: Facilitates real-time hazard reporting with integrated GPS location tracking and image submission capabilities.
• Administrative Dashboard: A comprehensive web interface for administrators and analysts to monitor, verify, and manage hazard reports.
• Backend API: A RESTful service built with Node.js and Express, utilizing MongoDB for data persistence and integrating AI models for automated verification.
• Secure Authentication: User identity management powered by Firebase Authentication.
• Role-Based Access Control: Granular permissions for different user roles, including administrators and data analysts.

Project Architecture

The repository is structured into three primary components:

• mobile-app: Developed using React Native and Expo for cross-platform mobile reporting.
• dashboard: A React.js based web application for administrative oversight.
• server: A Node.js and Express backend handling business logic and AI integration.

Technical Stack

• Frontend: React.js, React Native, Expo
• Backend: Node.js, Express
• Database: MongoDB
• Authentication: Firebase
• AI/ML: CLIP (openai/clip-vit-large-patch14) via Hugging Face
• Media Management: Cloudinary

---

Developed for SIH 2025 by Team 404 Founders in collaboration with INCOIS (Indian National Centre for Ocean Information Services).