Underwater Computer Vision Fish Biodiversity Estimation System

DCube developed an underwater computer vision–based system for automated fish biodiversity estimation using low-cost, non-invasive video sampling. The solution leverages custom detection and tracking algorithms to reliably monitor fish populations, supporting conservation efforts and sustainable management of Pakistan's aquatic resources.

The Client

Client Name: AI Fish Watch
Industry: Conservation
Region: Autralia
Company Size: Enterprise

The Challenge

Conventional fish population surveys are costly, invasive, and difficult to scale, often requiring specialized equipment and manual analysis. These methods can disturb marine ecosystems and are impractical for continuous biodiversity monitoring, particularly in resource-constrained environments.

The Solution

DCube designed a custom computer vision pipeline that processes underwater video footage to detect, track, and classify fish species in real time or batch mode. By combining bespoke detection and multi-object tracking algorithms optimized for underwater conditions, the system enables accurate, repeatable biodiversity estimation without physical interaction with marine life.

Key Features

• Custom computer vision–based fish detection algorithms
• Multi-object tracking to avoid double counting and improve accuracy
• Non-invasive, video-based sampling approach
• Optimized performance for underwater lighting and visibility challenges
• Support for low-cost camera hardware
• Scalable processing for large volumes of video data

Technologies Used

• Computer Vision
• Custom Detection & Tracking Algorithms
• Python

Results & Impact

• Enabled scalable, low-cost monitoring of fish biodiversity
• Reduced dependency on invasive and manual survey techniques
• Improved accuracy through tracking-based population estimation
• Supported preservation of endangered species
• Contributed to data-driven, sustainable management of Pakistan's fisheries