Visitors Have Fun Annotating Capybara Behavioral Data: átoa x Kobe Digital Lab Host First AI Experience Exhibit Allowing Participation in Aquarium Research (AI Ecological Observation Project Developed by Student Interns)

Kobe Digital Lab, Inc. (KDL; Chuo-ku, Kobe, Hyogo Prefecture; President: Shinichi Tamaki), a company specializing in AI-driven system development, is partnering with the theater-style aquarium 'átoa' (Chuo-ku, Kobe, Hyogo Prefecture; Director: Hiromi Nakayama) to host its first trial exhibit. This exhibit allows visitors to participate in a joint ecological observation project utilizing AI.

This exhibit is being held as a special project to celebrate the 5th anniversary of átoa's opening. It is a new type of interactive exhibit that allows visitors to experience 'annotation (*1),' a process necessary for AI learning, while simultaneously contributing to the aquarium's research activities.

AI-Powered Ecological Observation Pilot Project

At átoa, the focus extends beyond the care and display of animals to include ecological surveys and research. While the aquarium aimed to improve animal care environments and exhibits by analyzing animal behavior, a significant challenge remained: it is a heavy burden for keepers to constantly observe and record animal behavior alongside their daily duties.

Meanwhile, KDL has been hosting internships as part of its efforts to foster local talent and promote employment in Kobe, but faced the challenge of providing more engaging internship environments for students.

To solve these mutual challenges, the two parties began a joint pilot project in April 2023. Using data actually collected at the aquarium, students participating in KDL's internship program, under the guidance of KDL engineers, are developing AI and AI-based solutions to track, record, and visualize the behavior of captive animals.

In this pilot project, data is collected in a way that minimizes stress on the animals and impact on their environment. The capybara exhibit area is filmed by cameras installed on the ceiling, recording their daytime behavior as video. Based on this footage, AI models (*2) are being tuned to identify and track the position and movements of the capybaras in real-time. 'Annotation' is essential for this process.

Exhibit Overview

Dates: April 1, 2026 (Wed) – April 17, 2026 (Fri), 10:00 – 19:00 Location: Theater-style aquarium 'átoa' (Chuo-ku, Kobe, Hyogo Prefecture), 4th Floor SKYSHORE Participation Fee: Free (Separate aquarium admission required)

[Exhibit Details] In this exhibit, visitors can experience 'annotation' using touch panels installed in front of the capybara enclosure on the 4th floor of átoa.

Visitors can learn about how AI works and get a glimpse into research in a game-like, fun way. It is a rare initiative that allows visitors to participate in and contribute to an aquarium's research activities.

① Find the capybara in the real-time video of the enclosure displayed on the touch panel, and use touch-and-drag to draw a box around it.

② Select the behavior the capybara is performing from the provided options. Examples: 'Walking,' 'Swimming,' 'Rolling,' 'Running,' 'Sitting,' etc.

③ The data entered by visitors is displayed as aggregated results on a separate screen. Visitors can view a list of capybara behavior types and their distribution by time of day, allowing them to visually confirm how the data they entered is being accumulated.

Future Outlook

Since the start of the pilot project, KDL has been responsible for the design and construction of the AI models and the solutions that use them. The results obtained from this pilot project will also be applied to solving challenges in other fields using AI.

This exhibit also aims to expand the possibilities of data collection in animal behavior analysis. We will verify whether annotations made by visitors can be accumulated as behavioral analysis data and used to tune AI models. By utilizing this accumulated data, we aim to improve the accuracy of animal behavior analysis.

Internships Available

This pilot project is led by KDL interns. KDL is continuously recruiting interns to participate in this initiative. Please contact us for details.

KDL will continue to work on this pilot project as a partner that stands in the same position as the staff facing these challenges, thinking about and creating the possibilities of technology through ongoing dialogue.

*1: Annotation The process of processing data used for AI learning so that the AI can recognize it correctly. In object detection, the target object in an image is enclosed in a rectangle, and its type and attribute information are assigned (labeling). This creates a single piece of annotation data.

*2: AI Model A program that allows AI to analyze data and make predictions or judgments. It follows a flow where image data is 'input' into the model, the model 'analyzes' the input data, and 'outputs' results such as predictions or judgments based on that analysis.

[Related Press Release] Success in Digitizing and Visualizing Capybara Behavioral History Data with AI: Results of a Practical Internship Tackling Aquarium Challenges — A Collaboration between Kobe Digital Lab and Theater-style Aquarium 'átoa' (December 2024) https://www.kdl.co.jp/news/2024/12/news-6483/

[Company Overview] Company Name: Kobe Digital Lab, Inc. Representative: Shinichi Tamaki, President Address: Shin-Crescent Building, 72 Kyo-machi, Chuo-ku, Kobe, Hyogo 650-0034 Established: October 1995 Capital: 50 million yen Number of Employees: 172 (as of March 1, 2026) URL: https://www.kdl.co.jp/

[Contact for átoa] AQUARIUM x ART átoa Attn: Hirose, Ishihara, Handa TEL: 078-771-9393 FAX: 050-3156-3232 E-MAIL: press-atoa@aquament.co.jp

[Contact for Kobe Digital Lab] Kobe Digital Lab, Inc. Digital Business Division, Data Intelligence Team: Kohei Yamaguchi, Toshiki Haraguchi, Taichi Fukuoka E-mail: info@kdl.co.jp / Phone: 078-327-2280 *We are a company promoting telework. We would appreciate it if you could contact us via email.