Projects
Applied research projects in robotic perception, industrial vision, multimodal interaction, and reliable visual AI.
Hyundai–NTU–A*STAR Corporate Lab — Industrial Robotic Perception · Nov 2025 – Present
Building the perception stack for reliable robotic part picking at HMGICS: open-world object detection and segmentation, 6D pose estimation, and grasp pose generation for cluttered industrial scenes. Also developing foundation-model-based few-/zero-shot anomaly detection, adapting CLIP and DINOv3 to spot defects in unseen part categories without target-domain training data. [ACM MM 2026]
Schaeffler–NTU Corporate Lab, Phase III — Multimodal Human–Robot Collaboration · Jun 2025 – Nov 2025
Benchmarked multimodal 3D human pose estimation and developed motion-retargeting methods for human–robot collaborative tasks in industrial environments.
DSO National Laboratories — Robust Image Watermarking · Dec 2024 – Jun 2025
Built a robust encoder–decoder framework for image watermarking with tamper resistance under JPEG compression and noise. Proposed SSH-Net, a self-supervised hybrid network for noisy watermark removal without paired training data. [JVCIR 2025]
Schaeffler–NTU Corporate Lab, Phase II — Open-Vocabulary Industrial Vision · Mar 2024 – Dec 2024
Developed deep learning solutions for object detection and human–object interaction (HOI) recognition in industrial environments, including open-vocabulary methods and cross-level HOI distillation from multimodal LLMs. [KBS 2025, ACM MM 2025, EMNLP 2024]
National Research Foundation, Singapore (NRF) — Image and Video Restoration · Jan 2021 – Mar 2024
Designed learning-based frameworks to recover corrupted images and videos from flash storage. Work covers JPEG bitstream restoration, bitstream-corrupted video recovery, and multimedia file forensics. [CVPR 2023, NeurIPS 2023, IEEE TMM 2024, IEEE TMM 2026, DCC 2023]
Chongqing University — Video Decoding & AI Acceleration · Sep 2017 – Jul 2020
Researched energy-efficient video processing and machine learning acceleration. Developed task-level parallel H.264 decoding with dynamic voltage and frequency scaling, and contributed to HolyLight, a nanophotonic accelerator for data-center CNN workloads. [ICPADS 2018, DATE 2019]