BMVC 2026 · arXiv:2608.24763
MoTE: Mixture of Task Experts for Multi-Task Video Understanding
British Machine Vision Conference (BMVC), 2026.
MoTE converts decoder feed-forward networks into explicitly routed task experts while retaining a shared multimodal backbone, enabling interpretable and compute-efficient multi-task video-language learning.
VISAPP 2025 · arXiv:2608.20093
HandMvNet: Real-Time 3D Hand Pose Estimation Using Multi-View Cross-Attention Fusion
20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISAPP), 2025.
HandMvNet estimates absolute 3D hand joints and meshes from multiple RGB views through joint-aligned feature sampling and cross-attention fusion, with real-time calibrated and calibration-free variants.