BACK TO INDEX
CASE-2026-009 // JAL DRISHTI
RESTRICTEDAI Systems / Maritime Security

Jal Drishti

STATUS:ACTIVE
RISK:CRITICAL

EXECUTIVE SUMMARY

Jal Drishti is an AI-powered real-time underwater surveillance and anomaly detection system designed for maritime security. The system enhances degraded underwater visuals using GAN-based image restoration and performs real-time anomaly detection using lightweight object detection models, providing operators with actionable insights in low-visibility aquatic environments.

PROBLEM STATEMENT

Underwater environments severely degrade visual data due to turbidity, light absorption, and color distortion, making conventional computer vision systems unreliable. Jal Drishti addresses this challenge by restoring underwater image quality and enabling real-time anomaly detection to assist maritime operators in identifying potential threats or unusual underwater objects.

INVESTIGATIVE FINDINGS

  • ▸Underwater environments significantly reduce the accuracy of traditional computer vision models due to haze, scattering, and color attenuation
  • ▸Low visibility and inconsistent lighting conditions make manual underwater monitoring inefficient and error-prone
  • ▸Real-time maritime monitoring requires lightweight models capable of operating under strict latency constraints
  • ▸Combining image enhancement with object detection significantly improves detection reliability in degraded underwater footage

BEHAVIORAL ANALYSIS

Jal Drishti is designed as an assistive AI system for maritime operators. Instead of making autonomous decisions, it highlights potential anomalies and provides confidence-based alerts. This human-in-the-loop design ensures that AI supports situational awareness while allowing final judgment to remain with trained personnel.

TECHNICAL EXECUTION

The system processes incoming video streams frame-by-frame. Each frame undergoes enhancement using a lightweight FUnIE-GAN model to restore color balance, contrast, and clarity. The enhanced frames are then passed to a YOLOv8-Nano detector to identify anomalous underwater objects. A unified inference pipeline ensures correct normalization transitions between models and generates structured JSON outputs for visualization in a real-time dashboard.

SYSTEM ARCHITECTURE

Video Source → Backend Frame Scheduler → ML Inference Engine → GAN-Based Image Enhancement → YOLOv8-Nano Detection → Confidence & Safety Logic → JSON Result Output → Frontend Dashboard Visualization. The system architecture separates responsibilities across frontend, backend, and ML engine to maintain modularity, stability, and real-time performance.

OPERATIONAL FEATURES

  • ✓Real-time underwater video processing and enhancement
  • ✓GAN-based restoration of degraded underwater images
  • ✓Lightweight anomaly detection using YOLOv8-Nano
  • ✓Confidence-based threat classification system
  • ✓Split-screen operator dashboard showing raw and enhanced feeds
  • ✓WebSocket-based low-latency streaming pipeline

TECHNOLOGICAL ASSETS

ReactViteNode.jsFastAPIPyTorchYOLOv8-NanoFUnIE-GANWebSocket

CURRENT STATUS

Active development with completed phases including ML pipeline implementation, backend-frontend integration, and real-time prototype demonstration. Ongoing work focuses on system optimization, GPU acceleration, and operational stability.

CONCLUSION

Jal Drishti demonstrates the potential of combining underwater image enhancement with real-time anomaly detection to improve maritime situational awareness. By integrating AI-assisted perception with operator oversight, the system provides a practical prototype for future underwater surveillance and maritime security applications.

EVIDENCE LINKS

REPOSITORY

CASE METADATA

OPENED:2026-02-02
ID:CF-2026-009

TAGS

AI/Computer VisionMaritime SecurityUnderwater SurveillanceReal-Time SystemsPyTorchReact
SESSION ACTIVELAST SYNC: --:-- --/TRACE:HOME/CASES/CF-2026-009//ACTIVE[ESC] to exit contextreturning may invalidate findings
○ STANDARDCLEARANCE: LEVEL 2