Dongyue Lu
Lingdong Kong
Tianxin Huang
Gim Hee Lee
National University of Singapore
GEAL is a novel framework designed to enhance the generalization and robustness of 3D affordance learning by leveraging pre-trained 2D models. We employ a dual-branch architecture with Gaussian splatting to map 3D point clouds to 2D representations, enabling realistic renderings. Granularity-adaptive fusion and 2D-3D consistency alignment modules further strengthen cross-modal alignment and knowledge transfer, allowing the 3D branch to benefit from the rich semantics and generalization capacity of 2D models.
- [2024.12] - We have released our PIAD-C and LASO-C datasets on Hugging Face! 🎉📂
To be updated.