오션파크로 떠나는 즐거운 여행에 필요한 모든 것
Information technology and animal research experts at Ocean Park Hong Kong have adapted artificial intelligence (AI)-based people counting applications into advanced animal tracking systems, using machine learning to address the inherent human constraints in exhaustive and consistent behavioural data collection. These systems are now capable of real-time animal skeleton detection and pose recognition, thereby overcoming the limitations of past applications that could only detect animals as single entities in images.
“Ocean Park Hong Kong is committed to science-based animal husbandry and health management. By enhancing target detection algorithms and building customised AI models that go beyond whole object tasks to identify the body parts and pose of the Park’s signature species with high conservation value, we have gained substantial insights into individual animal behavioural patterns. This has added value for the team to provide more personalised animal care and habitat design. Also, understanding species-specific behaviours and interactions with the environment is crucial for research and effective conservation,” said Paulo Pong, Chairman of the Board of Ocean Park Corporation.
Gathering insights into the daily lives of giant pandas and golden monkeys
Ocean Park’s journey into machine learning assisted animal daily monitoring began in early 2025 with giant pandas An An and Ke Ke. This includes Ke Ke’s activity level and habitat use following modifications of habitat or the introduction of new enrichment.
The system has since been expanded to the golden monkey family, identifying individuals like the mother Le Le and father Qi Qi as adults and the daughter Little Red Bean as a juvenile, and meticulously tracking and analysing their unique movement patterns within their habitat, such as the distance changes between the baby and mother as the baby grows.
More precise and comprehensive data
The innovative system leverages machine learning to recognise key anatomical features of different species, for example, the giant panda’s head, limbs and body; and to identify a spectrum of animal behaviours, from simple distinctions like stationary versus moving, to intricate actions such as walking, climbing, rolling and jumping. The system detects specific movements – such as “raising the arms” (potentially a precursor to the motion of “rolling on the floor” for the giant pandas). This granular data provides critical insights for interpreting complex behavioural patterns, and building a comprehensive database for ongoing research, helping the team develop highly targeted animal care strategies.
Ocean Park’s industry-leading machine learning model significantly streamlines data analysis and recording, saving researchers countless hours previously dedicated to manually reviewing limited video footage. This shift empowers staff with unparalleled efficiency and accuracy, largely due to the model's ability to apply consistent standards. Moreover, the model excels at processing big data – massive, continuous datasets – a task impractical for manual methods. This capacity for extensive, 24-hour observation provides comprehensive data for robust comparative analysis by the Park’s expert teams, significantly enhancing data size and accuracy compared to the typical one-hour, twice-weekly manual coding.
Deeper analysis for customised care
The system can also generate detailed reports with charts, illustrating animal activities during specific time frames. Such data are highly useful for the research and animal caretaker teams. For instance, by understanding precisely where pandas spend the most time, the team can strategically place enrichment items elsewhere to encourage exploration and natural behaviours, and stimulate both their physical and mental health.
Looking ahead: AI for dolphin research and AI teaching package for students
Ocean Park Hong Kong is now preparing to embark on one of its most ambitious and technically challenging AI projects yet: the comprehensive monitoring of dolphins. The Park is committed to developing AI systems capable of analysing dolphins’ movements in a dynamic, multi-dimensional environment, and integrating acoustic data – paving the way for groundbreaking new standards in marine mammal care and research globally.
Ocean Park's application of artificial intelligence in animal behavioural research and husbandry care will also provide practical teaching content for Hong Kong schools, aligning with the Education Bureau’s drive to promote school-based AI education. Students will utilise real data and footage provided by Ocean Park to explore pre-trained AI models, and subsequently design their own small-scale AI research projects.
