Tran Dinh Phu
I am a Ph.D. student in Computer Science at the Korea Advanced Institute of Science and Technology (KAIST), advised by Prof. Daeyoung Kim in the DEAL Lab. I received my M.S. in Computer Science from KAIST (Feb 2025) under the same advisor, and my B.E. in Control Engineering & Automation from the Hanoi University of Science and Technology (HUST) in 2022, graduating in the top 1% of my program.
My research centers on low-level vision — image & video super-resolution, restoration, and enhancement — and efficient multimodal learning — audio-visual understanding and reasoning. I build systems that see, restore, and understand visual content at scale.
📄 Curriculum Vitae (PDF) · 📧 phutx2000@kaist.ac.kr · phutx2000@gmail.com
Research Interests
- Image & Video Restoration / Enhancement — efficient restoration architectures, diffusion-based generative super-resolution that addresses the trade-off between perceptual and fidelity quality, temporal consistency modeling, and foundation models for image and video enhancement.
- Vision–Audio Multimodal Learning — audio-visual question answering, long-form video understanding, multimodal reasoning, and agentic AI systems; parameter-efficient adaptation of foundation models, cross-modal representation learning, and scalable reasoning over long-horizon audio-video content.
- Model Generalization & Regularization — robust optimization with Sharpness-Aware Minimization (SAM), federated learning, and regularization techniques such as SWA for improving robustness, fairness, and training stability under limited data and compute budgets.
News
| Date | |
|---|---|
| Jun 2026 | 🎉 FreqOrtho-SR and MedCAGD accepted at ECCV 2026 (Rank A*) |
| May 2026 | 🎉 Efficient Encoder-Only Context Compression accepted at ICML 2026 Workshop |
| Feb 2026 | 🎉 SAT accepted at CVPR 2026 (Rank A*) |
| 2025 | 🏆 Awarded the Hyundai Global Scholarship for graduate study (5 slots/cycle) |
| Jun 2025 | 🎉 VSRM accepted at ICCV 2025 (Rank A*) |
| Feb 2025 | 🎓 Completed my M.S. in Computer Science at KAIST |
| Dec 2024 | 🎉 VieBookRead accepted at AAAI 2025 (Rank A*) |
| Jul 2024 | 🎉 CPAT accepted at BMVC 2024 (Rank A) |
Selected Publications
Full list on Google Scholar · Semantic Scholar · all publications. Underlined = me; * denotes equal contribution.
SAT: Selective Aggregation Transformer for Image Super-Resolution
Dinh Phu Tran, Thao Do, Saad Wazir, Seongah Kim, Seon Kwon Kim, Daeyoung Kim
IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), Finding Track, 2026
FreqOrtho-SR: Frequency-Guided Orthogonal Expert Learning for Real-World Image Super-Resolution
Minh Son Hoang*, Dinh Phu Tran*, Quyen Nguyen Duc, Dam Hoang Phuong, Daeyoung Kim
European Conf. on Computer Vision (ECCV), 2026
MedCAGD: Context-Aware Gated Decoder for Robust Medical Image Segmentation
Saad Wazir, Patrick Vibild, Dinh Phu Tran, Seongah Kim, Daeyoung Kim
European Conf. on Computer Vision (ECCV), 2026
VSRM: A Robust Mamba-Based Framework for Video Super-Resolution
Dinh Phu Tran, Dao Duy Hung, Daeyoung Kim
International Conf. on Computer Vision (ICCV), 2025
Reference-Based Post-OCR Processing with LLM for Precise Diacritic Text in Historical Document Recognition (VieBookRead)
Thao Do, Dinh Phu Tran, An Vo, Daeyoung Kim
AAAI Conf. on Artificial Intelligence (AAAI), Social Track, 2025
Channel-Partitioned Windowed Attention and Frequency Learning for Single Image Super-Resolution (CPAT)
Dinh Phu Tran, Dao Duy Hung, Daeyoung Kim
British Machine Vision Conference (BMVC), 2024
Also under review: Knowing When to Answer: Adaptive Confidence Refinement for Reliable Audio-Visual Question Answering and TB-AVA: Text as a Semantic Bridge for Audio-Visual Parameter Efficient Finetuning. Patent (under review): Video Super-Resolution Technique Leveraging Spatial-Temporal Mamba Architecture.
Honors & Awards
- Hyundai Global Scholarship for graduate study (5 slots per admission cycle) — 2025–2029
- KAIST Scholarship for the Ph.D. program (full tuition & fees) — 2025–2029
- KAIST Scholarship for the M.S. program (full tuition & fees) — 2023–2024
- Best Bachelor Thesis Award, HUST — 2022
- Scholarship Encouraging Study (Level A, highest) — 2020–2021, 2021–2022
- Nitori International Scholarship — 2019–2020
- High school valedictorian — 2018
Academic Service
Conference Reviewer — AAAI 2026, CVPR 2026, ICML 2026, NeurIPS 2026
Talks
- [KAIST, Fall 2025] Invited talk at CS632 Embedded Operating Systems — Super-Resolution Applications