Search papers, labs, and topics across Lattice.
1
0
3
Ditch global embeddings for text-motion retrieval: this method uses joint-angle motion images and token-patch late interaction to achieve state-of-the-art accuracy and interpretability.
