FLOWT in Action: AI-Powered Floating Litter Observation & Waste Tracking

FLOWT (Floating Litter Observation & Waste Tracking) is a modular AI-powered computer vision system designed to detect, monitor, and track plastic and floating debris in aquatic environments. The platform supports end-to-end video ingestion, object detection, visualisation, and tracking of individual debris items across video frames, generating reliable data to support clean-up efforts, policymaking, and long-term prevention.

FLOWT incorporates a human-in-the-loop review process that enables users to validate and correct detections, ensuring high-quality, trustworthy outputs for downstream analysis. The system operates through two complementary, yet independent workflows: Waste Tracking, which provides operational monitoring, annotated outputs, and analytical summaries without requiring machine learning expertise; and Model Improvement, which uses validated annotations to continuously retrain and fine-tune detection models while maintaining ongoing monitoring.

In addition, FLOWT integrates AI Insight, leveraging Large Language Models (LLMs) to interpret spatial and temporal patterns, flag anomalies, identify recurring errors, and support both operational decision-making and model refinement. By combining automated detection, expert validation, and AI-driven insights, FLOWT enables transparent, efficient, and continuously improving monitoring of floating debris.