National Oceanic Institute Marine Waste Image Identification Competition 6 countries duel Kuan Biling: Taiwan stands at the forefront

The National Oceanic Research Institute and Amazon Web Services (AWS), a subsidiary of Amazon, held the first international AI competition for marine waste image recognition, attracting 70 teams and 204 people from 6 countries including Taiwan, Indonesia, Vietnam, Haiti, Malaysia, and Ukraine to participate. An inter-school team from Tsinghua University and National Taiwan University won the championship with an image recognition rate of 81%. Chairman of the Ocean Council, Kuan Biling, presented the award today and said that Taiwan's first set of high-quality marine waste data collection has successfully moved from "AI-ready" to "AI practical implementation", and Taiwan has stood at the forefront of the marine AI field.

Guan Biling pointed out that the competition adopts the model of an international large-scale data science competition, with cloud computing power provided by AWS and real-time objective scoring combined with the ITRI AIdea platform, allowing teams to compete with each other based on their model identification capabilities under the same conditions. Marine AI is an emerging field for the world, and Taiwan has taken a leading position, and participating teams have also reached a more cutting-edge starting point.

Wang Dingkai, managing director of AWS Taiwan, said that many AI competitions lack real, high-quality data and the results are difficult to meet practical applications. The release of high-quality data sets that have been professionally labeled by the National Academy of Sciences, coupled with AWS's international cloud computing power, not only improves the practicality of the solution, but also gives Taiwan's results the opportunity to be used globally through the cloud.

Chen Zhangling, President of the China Academy of Ocean Sciences, said that the marine debris image data collection platform has been built for many years and contains more than 20,000 images and tens of thousands of items of marine debris. It has established a three-level structure based on the classification of international beach cleanup operations and is open with a CC BY 4.0 license for academic research and industry to invest in AI applications and import assessments into drones for automatic classification and statistics of coastal marine debris.

The championship was won by an inter-school team from Tsinghua University and National Taiwan University, and won the Best Lightweight Award; the runner-up was the team of Delta Electronics and Lun Guang Technology, the third runner-up was the team of National Central University and Taiwan Normal University, and the Hanxiang Aviation Industry Engineer Team stood out with its lightweight model.

The champion team said that by moving from the laboratory to real data, they deeply realized that the value of science and technology lies in responding to real needs; the runner-up team accurately identified tiny marine debris, even improving the accuracy to the fourth decimal place, demonstrating the practical potential of AI technology in ocean governance.