FOSK: Fast Open-Vocabulary 3D Instance Segmentation via Consensus-Filtered Knowledge Distillation
Published in ICASSP, 2026
Recommended citation: Wu, H., Gao, Z., Cao, J., Yao, K., Shen, W., & Wei, Z. FOSK: Fast Open-Vocabulary 3D Instance Segmentation via Consensus-Filtered Knowledge Distillation. In ICASSP 2026. https://ieeexplore.ieee.org/abstract/document/11461881/
Abstract. This paper proposes Fast Open-Vocabulary 3D Instance Segmentation via Consensus-Filtered Knowledge Distillation, which distills open-vocabulary knowledge from 2D embeddings into a 3D student model for faster inference.
Authors: Hongrui Wu, Zhicheng Gao, Jin Cao, Kelu Yao, Wen Shen, Zhihua Wei†.