Khanh Binh (Calvin) Nguyen

Associate Research Fellow
image

I am a dedicated AI Research Specialist at Deakin University, with expertise in Deep Learning, Computer Vision, and Multimodal AI systems. Holding a Ph.D. from Sungkyunkwan University, I spearhead innovative research in efficient neural architectures and advanced training methodologies, targeting applications in image recognition, few-shot learning, and Large Multimodal Language Models. My work, published in top-tier venues like CVPR, ICCV, and WACV, and recognized in global AI competitions, underscores my drive to deliver transformative AI solutions.

Professional Experience

Associate Research Fellow

Deakin University | Geelong, Australia | 2025 - Present
  • Developed state-of-the-art AI algorithms for real-time detection of invasive species using camera traps, bolstering ecological preservation.
  • Architected a video-based framework to identify endangered species, enhancing biodiversity monitoring with 95% accuracy.
  • Led cross-disciplinary teams to deploy scalable computer vision systems, bridging research and real-world impact.

Senior Deep Learning Engineer

Dnotitia | Seoul, South Korea | 2024 - 2025
  • Engineered multimodal AI frameworks with VAST, delivering ≥82% recall@5 for bilingual video retrieval, revolutionizing media search capabilities.
  • Curated the Korean Heritage Augmented Narrative (HAN) Dataset, advancing AI-driven cultural preservation.
  • Deployed GDPR-compliant, cloud-native AI solutions, reducing latency by 30% via model quantization and distributed training.
  • Pioneered few-shot learning techniques for low-resource languages, enabling rapid market scalability.

Medical AI Researcher

National Cancer Center | Goyang, South Korea | 2023 - 2024
  • Designed multimodal AI systems achieving ≈90% robustness for cancer diagnostics, enhancing patient outcomes.
  • Introduced SymMIM, a breakthrough Masked Image Modeling technique, accepted at ICPR 2024 for its novel symmetry-based approach.
  • Developed SAVE-AVS, an Audio-Visual Segmentation framework, elevating semantic segmentation precision with audio integration.
  • Adapted Segment Anything Model for brain MRI/CT analysis, achieving ≈90% Dice Similarity Coefficient across 7 categories.
  • Contributed as a peer reviewer for NeurIPS, CVPR, and AAAI, upholding rigorous standards in AI research.

Ph.D Researcher

DATES Lab, Yonsei University | 2019 - 2023

Specialized in Deep Neural Networks via supervised, self, and semi-supervised learning. Fully capable of the most up-to-date architectures in Image Recognition and optimized model efficiency via regularization techniques.

  • Invented Checkerboard Dropout, a regularization technique boosting CNN generalization, published in IEEE Access.
  • Created EUNConv and EUNNet, enabling efficient training of deep networks without batch normalization, streamlining workflows.
  • Formulated advanced semi-supervised learning pipelines (ReFixMatch, SequenceMatch), overcoming data bias challenges.
  • Reviewed for top conferences (CVPR, ICCV, NeurIPS), shaping the direction of AI innovation.

Publications

Recent Publications

  • Retro: Reusing teacher projection head for efficient embedding distillation on Lightweight Models via Self-supervised Learning [link]

    • Khanh-Binh Nguyen, Chae Jung Park
    IEEE Access, 2025

  • On Calibration of Prompt Learning Using Temperature Scaling [link]

    • Khanh-Binh Nguyen, Chae Jung Park
    IEEE Access, 2025

  • SAVE: Segment Audio-Visual Easy way using Segment Anything Model [link]

    • Khanh-Binh Nguyen, Chae Jung Park
    Computer Vision and Image Understanding (CVIU), 2025

  • Souple: Enhancing Audio-Visual Localization and Segmentation with Learnable Prompt Contexts [link]

    • Khanh-Binh Nguyen, Chae Jung Park
    Computer Vision and Image Understanding (CVIU), 2025

  • Symmetric masking strategy enhances the performance of Masked Image Modeling [link]

    • Khanh-Binh Nguyen, Chae Jung Park
    The British Machine Vision Conference (BMVC), 2024

  • SequenceMatch: Revisiting the design of weak-strong augmentations for Semi-supervised learning [link]

    • Khanh-Binh Nguyen
    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2024 (Oral, top 3%)

  • Debiasing, calibrating, and improving Semi-supervised Learning performance via simple Ensemble Projector [link]

    • Khanh-Binh Nguyen
    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2024

  • Boosting Semi-Supervised Learning by bridging high and low-confidence predictions [link]

    • Khanh-Binh Nguyen and Joon-Sung Yang
    ICCV Workshop, 2023

  • EUNNet: Efficient UN-normalized Convolution layer for stable training of Deep Residual Networks without Batch Normalization layer [link]

    • Khanh-Binh Nguyen, Jaehyuk Choi, and Joon-Sung Yang
    IEEE Access, 2023

  • Checkerboard Dropout: A Structured Dropout With Checkerboard Pattern for Convolutional Neural Networks [link]

    • Khanh-Binh Nguyen, Jaehyuk Choi, and Joon-Sung Yang
    IEEE Access, 2022