Search papers, labs, and topics across Lattice.
3
0
3
7
BRepCLIP outperforms traditional point-based methods by 40% in CAD model retrieval, showcasing the power of structure-aware pretraining.
Text-to-3D models can now learn new object categories without forgetting old ones, thanks to a replay-based continual learning framework.
Knowledge distillation can make state-of-the-art object detectors practical for real-world deployment on edge devices, but the landscape of KD techniques is complex and rapidly evolving.