Damage Identification in Social Media Posts using Multimodal Deep Learning: code and dataset

This project aims at predicting natural disasters using Machine Learning. This project was submitted as a part of Code.Fun.Do++ hackathon organised by Microsoft in 2018.

Climate Disaster Warning System is a deep learning-based project for detecting wildfires, floods, and sea-level rise using satellite and ground data. It leverages ResNet, Vision Transformer (ViT), and GRACE datasets to support early warning systems and climate research.

Disaster Tweets API - Production-ready BERT-based FastAPI service for classifying tweets as Disaster or Not Disaster, with Docker, CI, and cloud deployment support.

Alert Text Detector is an NLP-based model that detects alert messages from social media posts. It is built using BERTweet Base and trained on a dataset of 23,000 tweets (alert & non-alert). The model flags emergency-related messages and classifies tweets based on textual content.

Alert Text Detector is an NLP-based model that detects alert messages from social media posts. It is built using BERTweet Base and trained on a dataset of 23,000 tweets (alert & non-alert). The model flags emergency-related messages and classifies tweets based on textual content.

LLL-based Disaster Detector Agentic AI Application : This project enables the detection and interpretation of environmental threats (e.g., floods, infrastructure risks) by leveraging large language models (LLMs) and multimodal inputs derived from CCTV-based river surveillance feeds.

Implementation of a Deep Neural Architecture to perform real-time semantic segmentation of forest fires in aerial imagery captured by drones.

NLPrescue is an advanced Natural Language Processing system designed to detect and classify disaster-related tweets in real-time. Built with PyTorch and modern NLP techniques, it helps emergency responders quickly identify genuine disaster situations on social media platforms.