A5030285894- Domain Adaptation and Few-Shot Learning
- Multimodal Machine Learning Applications
- Advanced Neural Network Applications
- Remote Sensing and LiDAR Applications
- Synthetic Aperture Radar (SAR) Applications and Techniques
- Soil Moisture and Remote Sensing
- Autonomous Vehicle Technology and Safety
- Human Pose and Action Recognition
- Biomedical Text Mining and Ontologies
- Topic Modeling
- 3D Surveying and Cultural Heritage
- Artificial Intelligence in Healthcare and Education
Taiwan Forestry Research Institute
2022-2025
National Central University
2021-2024
The reported study illustrates a methodology for compiling training datasets to fine-tune Large Language Models (LLMs) healthcare information in Vietnamese, low-resource language. objective is bridge the gap medical accessibility and enhance communication developing countries by adapting LLMs specific linguistic nuances domain needs. involves selecting base model, domain-specific dataset, fine-tuning model with this dataset. Three open-source models were selected. comprising approximately...
Recently, self-supervised learning methods have been shown to be very powerful and efficient for yielding robust representation by maximizing the similarity across different augmented views in embedding vector space. However, main challenge is generating with random cropping; semantic feature might exist differently leading inappropriately objective. We tackle this problem introducing Heuristic Attention Representation Learning (HARL). This framework relies on joint architecture which two...
Affordance understanding, the task of identifying actionable regions on 3D objects, plays a vital role in allowing robotic systems to engage with and operate within physical world. Although Visual Language Models (VLMs) have excelled high-level reasoning long-horizon planning for manipulation, they still fall short grasping nuanced properties required effective human-robot interaction. In this paper, we introduce PAVLM (Point cloud Vision-Language Model), an innovative framework that...
In recent years, self-supervised learning has been studied to deal with the limitation of available labeled-dataset. Among major components learning, data augmentation pipeline is one key factor in enhancing resulting performance. However, most researchers manually designed pipeline, and limited collections transformation may cause lack robustness learned feature representation. this work, we proposed Multi-Augmentations for Self-Supervised Representation Learning (MA-SSRL), which fully...
Recognizing the trend and mechanism of land subsidence for reducing impact is critical in Taiwan. However, region western Taiwan consists complex aquifers/aquitards, requiring multilevel measurement techniques profiling. The commonly practiced way multi-leveling with magnetic ring installation, but suffering apparent error less temporal resolution due to manual operation. This study adopted Time Domain Reflectometry (TDR) as core technology automatic high accuracy instrumentation solve...