Open-Vocabulary Industrial Vision (Schaeffler Phase II)
Deep learning for object detection and human–object interaction in industrial environments. Schaeffler–NTU Corporate Lab · Mar 2024 – Dec 2024
Funding: Schaeffler–NTU Corporate Lab · Mar 2024 – Dec 2024
Developed deep learning solutions for object detection and human–object interaction (HOI) in industrial environments. The work focused on building robust visual perception systems that generalise to novel factory scenarios with limited supervision.
Contributions:
- Developed open-vocabulary detection and segmentation methods using language prompts and visual references for industrial scene understanding.
- Studied cross-level HOI recognition via knowledge distillation from multimodal large language models (KBS 2025).
- Contributed to continual learning methods for driver action recognition (ICIP 2024) and video question answering (EMNLP 2024).
- Co-authored grounded situation recognition using foundation model distillation at semantic, scene, and instance levels (ACM MM 2025).
Related publications: KBS 2025 · ACM MM 2025 · EMNLP 2024 · ICIP 2024