MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641051066 A) filed by Rithik V Kumar; Jeston Abraham D; and Sam Varghese George on April 22, 2026, for Gan-Augmented Attention Network For Real-Time Underwater Mine And Anomaly Detection.
Inventor includes Sam Varghese George.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The present invention proposes a real-time undetwater detection system using the "GANAugmented Attention Network for Real-Time Undetwater Mine and Anomaly Detection" framework, designed to accurately identify objects in complex maritime environments. During the development phase, the system utilizes a FUniE-GAN style U-Net generator as a data augmentation tool to synthesize high-fidelity training data and balance a strict four-class military taxonomy, effectively overcoming undetwater data scarcity and light attenuation issues. For realtime operations, the framework integrates a CBAM directly into a YOLOv8 backbone, creating a learnable feature refinement block that focuses on spatial and channel-specific details while suppressing background silt and noise. This integrated approach replaces traditional fixedpreprocessing methods, allowing the model to adapt to dynamic water conditions and achieve a low 4% false-positive rate. During inference, the system processes raw visual data to generate precise bounding boxes and confidence scores, which are presented through an intuitive dashboard interface. The resulting architecture provides a robust, low-latency solution suitable for marine monitoring, environmental assessment, and autonomous undetwater vehicle (AUV) operations.
Disclaimer: Curated by HT Syndication.
