Publications

2025

1. ICISN 2025

   

   An Automated Pipeline for Constructing a Vietnamese VQA-NLE Dataset

   Truong-Binh Duong, Hoang-Minh Tran, Binh-Nam Le-Nguyen, and 1 more author

   In Proceedings of the Fifth International Conference on Intelligent Systems and Networks, 2025

   Abs DOI Bib HTML PDF

   We present ViVQA-X, the first Vietnamese Visual Question Answering dataset with natural language explanations. Our work introduces an automated pipeline leveraging multiple Large Language Models as evaluators to construct a high-quality Vietnamese VQA dataset with explanations.

   @inproceedings{duong2026vivqax,
     title = {An Automated Pipeline for Constructing a Vietnamese VQA-NLE Dataset},
     author = {Duong, Truong-Binh and Tran, Hoang-Minh and Le-Nguyen, Binh-Nam and Duong, Dinh-Thang},
     booktitle = {Proceedings of the Fifth International Conference on Intelligent Systems and Networks},
     series = {Lecture Notes in Networks and Systems},
     pages = {164--173},
     year = {2025},
     publisher = {Springer Nature Singapore},
     isbn = {978-981-95-1746-6},
     doi = {10.1007/978-981-95-1746-6_18},
     github = {duongtruongbinh/ViVQA-X},
   }

2. ICCV Workshop

   

   Describe Anything Model for Visual Question Answering on Text-rich Images

   Yen-Linh Vu^*, Dinh-Thang Duong^*, Truong-Binh Duong, and 1 more author

   ICCV Workshop, 2025

   *Equal contribution

   Abs arXiv Bib HTML

   We propose the Describe Anything Model (DAM-QA) for Visual Question Answering on text-rich images. Our approach demonstrates strong performance across six benchmark datasets through comprehensive evaluation of multiple Vision-Language Models.

   @article{vu2025dam,
     title = {Describe Anything Model for Visual Question Answering on Text-rich Images},
     author = {Vu, Yen-Linh and Duong, Dinh-Thang and Duong, Truong-Binh and others},
     journal = {ICCV Workshop},
     year = {2025},
     publisher = {IEEE},
     github = {Linvyl/DAM-QA},
     note = {*Equal contribution}
   }

2024

1. HackHCMC

   

   Heineken Image Analysis Tool: Multi-Model AI Pipeline for Brand Compliance

   Truong-Binh Duong and Team AIO_CGLG2.0

   In HackHCMC 2024, Jul 2024

   Hackathon Project

   Abs HTML

   Engineered a multi-model AI pipeline integrating YOLOv10, Owlv2, PaddleOCR, and CLIP for automated brand compliance and safety analysis from images. Deployed using FastAPI backend for robust model inference.

2023

1. Personal

   

   Math Solving Assistant: AI-Powered Mathematical Problem Solver

   Truong-Binh Duong

   Apr 2023

   Personal Project

   Abs HTML

   Developed a chatbot website to solve math problems using Gemini API, integrating OCR for image input. Fine-tuned Gemini on MetaMathQA-40K with Vertex AI for improved mathematical reasoning.

2022

1. Course

   

   Manga Popularity Prediction: End-to-End Data Science Pipeline

   Truong-Binh Duong and Team

   Apr 2022

   Course Project

   Abs HTML

   Conducted comprehensive data science pipeline including web scraping, exploratory analysis, and predictive modeling. Implemented multiple regression models (Linear Regression, Random Forest, XGBoost) with Bayesian optimization for hyperparameter tuning.