Sustainable Development Goals

News Content

November 12, 2026 | 5:52 PM

By Hung Ting-hung, Kaohsiung

The National Academy of Marine Research (NAMR), in collaboration with Amazon Web Services (AWS), has hosted the first International AI Competition for Marine Debris Image Recognition, attracting 70 teams and 204 participants from six countries: Taiwan, Indonesia, Vietnam, Haiti, Malaysia, and Ukraine. A cross-university team from NTHU and National Taiwan University (NTU) won first place with an image recognition accuracy rate of 81%. At the awards ceremony today, Ocean Affairs Council Minister Kuan Bi-ling stated that Taiwan's first high-quality marine debris dataset has successfully progressed from being "AI-ready" to practical AI implementation, demonstrating that Taiwan is at the forefront of marine AI technology.

Kuan noted that the competition adopted the format of major international data science competitions. AWS provided cloud computing resources, while the Industrial Technology Research Institute's (ITRI) AIdea platform facilitated real-time, objective scoring. This enabled all participating teams to compete on equal terms based on their models' recognition capabilities. Marine AI remains an emerging field worldwide, and Taiwan has already established a leading position, giving participating teams an opportunity to compete at the cutting edge of the field.

Wang Ting-kai, Managing Director of AWS Taiwan, stated that many AI competitions struggle to produce practical results because of a lack of authentic, high-quality datasets. The NAMR's release of a high-quality dataset professionally annotated by human experts, combined with AWS's global cloud computing capabilities, not only enhances the practical applicability of AI solutions but also creates opportunities for Taiwan's innovations to be accessed and utilized worldwide through the cloud.

NAMR President Chen Chang-ling explained that the marine debris image dataset platform took several years to develop. It contains more than 20,000 images and tens of thousands of marine debris objects, organized into a three-level classification system based on the International Coastal Cleanup (ICC) classification framework. The dataset is openly available under the CC BY 4.0 license, allowing academic institutions, research organizations, and industry to develop AI applications. NAMR will also evaluate the use of drones for automated classification and statistical analysis of marine debris along coastlines.

The championship was won by the cross-university team from NTHU and NTU, which also received the Best Lightweight Model Award. The team from Delta Electronics and Luminance Technology took second place, while the joint team from National Central University and National Taiwan Normal University secured third place. Engineers from Aerospace Industrial Development Corporation (AIDC) also stood out with their lightweight AI model.

The championship team said that moving from laboratory experiments to real-world data gave them a deeper understanding that the true value of technology lies in addressing practical needs. The runner-up team, meanwhile, pursued even the smallest improvements in precision when identifying tiny pieces of marine debris, paying attention to accuracy improvements down to the fourth decimal place. Their efforts demonstrated the practical potential of applying AI technology to marine governance.

References