SeaScan: AI Fish Detection & Tracking
Real-time underwater fish detection and tracking system running YOLOv8 and NvDCF at 2K 60FPS on NVIDIA Jetson Orin for the sportfishing industry.
The Challenge
Traditional sportfishing methods require constant visual monitoring of underwater areas. The client needed an automated, AI-powered system that could detect and track fish in real-time at high resolution, distinguishing them from bait and other underwater objects.
Our Approach
We built SeaScan — a cutting-edge AI system combining deep learning detection with real-time tracking on edge hardware.
Deep Learning Detection
YOLOv8 architecture trained on a diverse dataset of underwater fish images ensures instantaneous and accurate identification. Proprietary algorithms distinguish fish from bait using size, shape, and movement patterns.
Real-Time Tracking
NvDCF tracking maintains continuous tracking of individual fish with high precision across frames, enabling users to follow specific targets.
Edge Processing
NVIDIA Jetson Orin 64GB provides compact, powerful computing. The NVIDIA DeepStream framework ensures efficient processing at full 2K resolution and 60 FPS.
High-Resolution Capture
A 2K 60 FPS camera ensures the detection and tracking stream is both displayed and saved in full resolution, providing unprecedented underwater visibility.
User Interface
Built in C++ with Qt, the interface optimizes hardware usage while providing a sophisticated, user-friendly experience for real-time monitoring.
Technical Stack
- AI: YOLOv8, NvDCF Tracking, NVIDIA DeepStream
- Hardware: NVIDIA Jetson Orin 64GB, 2K 60 FPS Camera
- UI: C++, Qt Framework
- Algorithms: Proprietary fish-bait distinction
Results
- Real-time fish detection and tracking at 2K 60 FPS
- Automated AI-powered system eliminating constant visual monitoring
- High-accuracy fish-bait distinction using proprietary algorithms
- Breakthrough product positioning in the sportfishing market