Revolutionizing Marine Conservation with AI: The Fight Against Ghost Nets
In a monumental step toward preserving marine life, conservationists have unveiled GhostNetZero.ai, an innovative artificial intelligence (AI) tool designed to detect abandoned fishing nets, commonly referred to as ghost nets. This groundbreaking technology has the potential to sift through an immense volume of underwater imaging data sourced from various organizations across the globe, identifying these hazardous nets and pinpointing their exact locations for removal.
Every year, an estimated 2% of the world’s fishing gear, which equates to around 80,000 square kilometers of fishing nets, is lost to the sea. These ghost nets pose a significant threat to marine wildlife, leading to the entanglement and drowning of seals, turtles, and dolphins. In addition, these abandoned fishing instruments decompose in the ocean, exacerbating plastic pollution—a growing environmental crisis.
Identifying ghost nets on the ocean floor is no small feat. Their fine, nearly invisible filaments make them incredibly difficult to locate without the aid of advanced technology. With GhostNetZero.ai, the collaboration between WWF Germany, Accenture, and Microsoft’s AI for Good Lab promises an effective solution by leveraging high-resolution side scan sonar images gathered by research institutes, government bodies, and offshore wind-power companies.
At the core of GhostNetZero.ai lies an advanced Convolutional Neural Network (CNN) that employs DeepLabV3 with a ResNet50 backbone. This sophisticated AI system has proven to be highly efficient, successfully identifying ghost nets in sonar scans with an impressive accuracy rate of 94%. As Gabriele Dederer, WWF Germany’s ghost nets project manager, notes, “Every side scan sonar image comes with geolocation and metadata, enabling the AI to not only spot a ghost net but also provide precise location details.”
Catching Ghost Nets
Once the AI flags a potential ghost net, the next vital step involves collaboration with local divers and fishermen to confirm the network’s presence. Dederer and her team seek local partners to verify AI findings, gathering essential details, such as approximate sizes of the nets. The logistics of retrieving these nets is a cooperative effort involving local maritime and conservation groups.
“The expertise of divers is indispensable. They can assess whether we’re dealing with large nets or smaller ones, which in turn helps determine the type and size of the retrieval ship required,” explains Dederer. The task of removing ghost nets is not just an environmental challenge; it requires significant logistical planning and financial investment, emphasizing the importance of a region-specific approach to the issue.
Currently, WWF Germany has established partnerships with local teams in France, Estonia, and Sweden, with plans to broaden its collaboration network in the future. This community-centric approach enhances the effectiveness of the operation while fostering local engagement in marine conservation efforts.
This initiative traces back to Dederer’s inspiration for utilizing crowdsourced sonar data, a vision that recently gained traction due to the support and technological prowess offered by Accenture and Microsoft. The integration of AI into the project took place earlier this year, marking a significant advancement in its capabilities.
The CNN, hosted on Microsoft Azure cloud infrastructure, utilizes PyTorch libraries and NVIDIA A100 TensorCore GPUs for efficient training and inference. According to Christian Bucher, Microsoft’s liaison for the ghost net project, the synergy between AI and local knowledge is instrumental. “On-the-ground verification contributes valuable insights, which are fed back into the model to enhance accuracy and efficacy,” he highlights.
“AI excels at pattern recognition, and by leveraging PyTorch, we can detect different types of ghost nets. Sonar images tend to present variations, making machine learning a fitting solution,” Bucher adds, reflecting on the approach that has led to achieving a 94% accuracy rate with the AI tool.
The GhostNetZero initiative sheds light on a pressing environmental challenge while harnessing the power of cutting-edge technology to facilitate remarkable strides in ocean conservation. Through collaboration and innovation, efforts are underway to mitigate the detrimental impact of ghost nets on marine ecosystems.
Inspired by: Source
