maritime-surveillance

📡 Production-grade maritime surveillance system for detecting ships in Synthetic Aperture Radar (SAR) imagery using YOLOv8-OBB (Oriented Bounding Boxes) and Lee speckle filtering. Handles the unique challenges of radar imagery: multiplicative noise and arbitrarily-oriented vessels.

YOLOv8 SAR Ship Detection leverages the YOLOv8 model to detect ships in SAR images and videos. It provides a Streamlit app for real-time detection, displaying bounding boxes around ships with or without confidence scores.

A unified tool for processing various SAR (Synthetic Aperture Radar) ship detection datasets into a standardized format. This tool supports multiple popular datasets including HRSID, SAR-Ship-Dataset, SSDD, and SRSDD-V1.0, with robust error handling and comprehensive logging.

SentinelBlue is an edge-deployed deep learning system for autonomous maritime search and rescue, featuring real-time RGB and thermal fusion for robust detection of distressed individuals and watercraft from UAV platforms.

Global Naval Surveillance System for Satellite Imagery

MACOW: Maritime Collision Warning System

AI-powered maritime surveillance system that detects ships and oil spills from SAR satellite imagery in real time. Built with TransUNet, Deformable DETR, and a gigapixel image viewer for human-in-the-loop validation.

Code release for EnvShip-Bench, a vessel trajectory benchmark built from DMA and NOAA AIS data, with compact, environment- aware, and interaction-aware subsets.

🏆 6th of 64 at Claude Hackathon, Imperial College London. AI-powered maritime surveillance fusing satellite AIS tracking, sanctions lists, RFMO registries and MPA boundaries into a 7-signal risk score, with Claude generating real-time intelligence briefs on illegal fishing vessels.

Dark vessel detection: fuse Sentinel-1 SAR imagery with live AIS to find ships that are not broadcasting — with ship-size estimation and a human-in-the-loop review dashboard

Industrial-grade SAR object detection using PyTorch

Detecting undeclared vessels with deep learning on Sentinel-1 SAR, fused with AIS over Danish waters

AI-powered ship detection system using Faster R-CNN and PyTorch to detect ships from range-compressed airborne radar (SAR) images for maritime surveillance.

Maritime surveillance software using AIS data (Automatic Identification System) for identifying illegal fishing transshipments. IMTA School Project in collaboration with Cerema

Self-supervised maritime anomaly detection using Transformer representation learning, HNSW behavioral retrieval, and multimodal AIS trajectory intelligence.

Satellite vessel detection & port activity monitoring from Sentinel-1 SAR. Any port, any time. Python package on PyPI. Validated on Singapore (R²=0.478 for bunker sales).

World Models for Maritime Mobility

Search theory for wide-area ISR: sweep width, coverage rate, Koopman random search, cross-cueing, track persistence and operator false-alarm burden. Tested Python with a CLI that writes POD-versus-time curves.

Autonomous Satellite Radar Oil Slick Intelligence & Sovereign Polluter Attribution Platform (Modern C++20 | Copernicus Sentinel-1 C-Band SAR)

Detection and monitoring of non-broadcasting "dark fleet" vessels using Synthetic Aperture Radar (SAR) imagery and deep learning, cross-referenced with AIS data for maritime surveillance.