Public dataset·Cross-modal learning
ObjectFolder 2.0
Multisensory neural objects with visual, acoustic, and tactile signals.
About this dataset
A dataset of 1,000 objects encoded as implicit neural representations that render vision, sound, and touch on demand. Ideal for pretraining cross-modal models and generating tactile readings at arbitrary contact points.
Citation
Gao et al. “ObjectFolder 2.0.” CVPR 2022.