Personal Pages

from world import netizens
print("Hello world!")
print("Greetings from Borneo Islands...!")

Research Agendas

We have experienced the outstanding performance of artificial intelligence surpassing human capability. Yet, they are unable to acknowledge fully that they do not know and learn the novel phenomenon continuously like humans. In my current research agenda, I am exploring a technique to enhance an agent to identify unknown categories and learn them continuously in domain-specific problem. This will make the agent become reliable and able to adapt to the open-world environment in the agent interest only. The applications can be applied to ecology, healthcare, smart cities, robotics, chatbot or intelligent systems. We named this as Domain-Specific Open-World Recognition.

We are interested in the domain-specific open-world recognition problems in Environmental/Biodiversity Monitoring and Conservation, Lifelong Chatbot, AI-generated content detection, and Agriculture/Food.

Short Biography

Hi, greetings from Borneo Island! I am an assistant professor at Institut Teknologi Kalimantan at the Department of Informatics. I have been teaching several topics including algorithms and programming languages, data structures, functional programming, numerical methods, machine learning, artificial intelligence, deep learning, web intelligence, and software engineering. I have reviewed several articles in Springer and Elsevier Journals. My research interest includes:

• Open-World Lifelong Machine Learning (OWLML)
• Ecological and Environmental Informatics
• Chatbot with OWLML

Main Projects

Domain-Specific Open-World Lifelong Artificial Intelligence

Current AI models demonstrate outstanding capabilities, yet they often struggle in non-stationary environments. This limitation can lead to unreliable outputs or hallucinated predictions. Therefore, developing AI systems that can adapt to continuously changing environments, especially for domain-specific tasks, is crucial for improving efficiency and achieving more targeted outcomes. In this project, we explore AI models that integrate open-set recognition, active learning, continual learning, and novel class discovery across different modalities, such as vision and text.

Students Ongoing projects

1. Open-Set Recognition for AI-Generated Image Detection
2. Open-Set Recognition for Palm Oil Disease Detection

Open-World Lifelong Learning for Biodiversity Monitoring

In a world teeming with diverse ecosystems, continuous and adaptive biodiversity monitoring is vital, as biodiversity plays a crucial role in our sustainability as humans. Open-world lifelong learning offers an innovative approach to monitoring that evolves alongside the environment. Unlike traditional models, which require retraining with new data and struggle to identify unknown classes, open-world lifelong learning systems autonomously learn and adapt over time, recognizing and integrating new species and ecological changes without restarting from scratch. The primary challenge lies in mitigating catastrophic interference to achieve true open-world capability.

Ongoing research

1. Open-Set recognition with continual learning capability for plant species classification

Students Ongoing projects

1. Active Learning Strategies for Plant Species Classification
2. An Integrated Biodiversity and Forestry Portal for Borneo
3. Large-Scale Open-Set Plant-Species Recognition using Lightweight Deep-Learning Models