CV

AI Instructor

MCI Academy

Developing and instructing AI Engineer / AI Application / n8n courses

AI Engineer

VNGGames

Working on AI Play Studio, research and development of AI tools & products for game development

Assistant Researcher

Visual Computing Lab - SeoulTech

Working on Korean government projects and writing a research paper

Software Engineer

Bosch Global Software Technologies

Working as a Software Developer for projects from China and Japan. Task: unit testing, integration testing, component/function development, devops, lab testing, static code analysis,...

Master Program

Seoul National University of Science and Technology

Computer Vision - 33 credits in 4 semesters

• Introduction of Unmanned Robotic Vehicles
• Reinforcement Learning
• Machine Learning
• Statistic Machine Learning
• Optimization Algorithm
• Digital Image Processing
• Advanced Topics on Information Technology
• Advanced Natural Language Processing
• Introduction to Deep Learning
• ...

Master Program

Ho Chi Minh City University of Technology

Data Science - 18 credits in 1 semester

• Programming Foundation for Data Analysis and Visualization
• Advanced Algorithm
• Mathematical Foundation for Computer Science
• Methodology of Scientific Research
• Business Ethics and Corporate Social Responsibility
• Philosophy

Excellent Engineer Program

Ho Chi Minh City University of Technology

Mechatronics - 268 credits in 5 years

• Statistic Methods and Data Analysis
• Robotics
• Intelligent Actuators
• Mechatronics System Design
• Digital Signal Processing and Application
• Object Oriented Design and Analysis
• Advanced Programming Language
• Advanced Data Structure
• Numerical Analysis and Optimization
• Linear and Nonlinear Control Systems
• ...

Scholarship for PhD Program

Full Scholarship

100% tuition fee + living expenses for 4 PhD years at Seoul National University of Science and Technology in South Korea

Scholarship for Master Program

Tuition scholarships

75% tuition scholarship for all semesters at Seoul National University of Science and Technology in South Korea

Scholarship for Master Program

Korean Professor

Full scholarship for studying and researching at Visual Computing Lab of Seoul National University of Science and Technology in South Korea

Final round for Exchange Program

Pasona Global

Full scholarship for students all over Vietnam. Reached the final round (Top 5) of the Pasona International Exchange Program for students doing summer internships in Japan

Scholarship for Engineer Program

Ho Chi Minh City University of Technology

Full scholarship to learn Japanese to prepare to work as an engineer in Japan. The only student who does not need to study Japanese language preparation for the scholarship.

Scholarship for Undergraduate Program

Tuition scholarships

Full scholarship for students all over Vietnam. Top 1/2 students received 100% tuition scholarships for Business Administration at International Pacific University in Japan

Final round for Undergraduate Program

Mitsubishi Heavy Industries

Full scholarship for students all over Vietnam to study Japanese and university in Japan. Reached the final round (Top 6) of the scholarship

Scholarship for Undergraduate Program

Vietnamese Government

Full scholarships for 10 students from all over Vietnam to study nuclear energy in Japan

SP2Mask4D: Efficient 4D Panoptic Segmentation Using Superpoint Transformers

IEEE

The increasing need for precise segmentation in dynamic outdoor environments, particularly with LiDAR data, has brought attention to the 4D panoptic segmentation task. This task requires accurate identification of both objects and semantic labels across spatial and temporal dimensions. In this work, we present SP2Mask4D, a novel approach that replaces the commonly used transformer architecture with a superpoint-based transformer architecture. This modification leads to faster inference and reduced memory consumption, while maintaining competitive performance compared to transformer-based methods. While both approaches use attention mechanisms, traditional transformer models apply attention to all points, resulting in high computational costs. In contrast, SP2Mask4D focuses attention within localized superpoints, significantly lowering the computational burden. Experiments on the SemanticKITTI dataset show that SP2Mask4D reduces inference time by about 32.8% and improves memory efficiency by 60.3%, while preserving segmentation performance comparable to state-of-the-art methods.

Transformer-supported Tackling 3D Point Cloud Instance Segmentation

ACVYS 2024

This study aims to address these challenges by utilizing the strengths of transformer models, a recent innovation in machine learning known for their ability to handle large-scale data efficiently. By reducing the number of point clouds required for training, the approach simplifies the overall model architecture and removes unnecessary intermediate steps. This not only decreases training time but also significantly reduces model complexity, making the process more streamlined and resource-efficient. Despite these reductions, the method maintains high accuracy in classifying and recognizing objects in 3D environments, ensuring robust performance even in dynamic and cluttered scenarios commonly encountered in autonomous driving. Thus, this research presents a more effective and efficient approach to 3D point cloud instance segmentation, pushing the boundaries of what is possible in AI-driven 3D object recognition.

MSTA3D: Multi-scale Twin-Attention for 3D Instance Segmentation

ACM Multimedia 2024

Recently, transformer-based techniques incorporating superpoints have become prevalent in 3D instance segmentation. However, they often encounter an over-segmentation problem, especially noticeable with large objects. Additionally, unreliable mask predictions stemming from superpoint mask prediction further compound this issue. To address these challenges, we propose a novel framework called MSTA3D. It leverages multi-scale feature representation and introduces a twin-attention mechanism to effectively capture them. Furthermore, MSTA3D integrates a box query with a box regularizer, offering a complementary spatial constraint alongside semantic queries.