Search papers, labs, and topics across Lattice.
Affiliation:
3
0
2
Hallucination-driven error propagation is tackled head-on, enabling state-of-the-art segmentation performance in adverse weather conditions through innovative cross-task collaboration.
Vision-language models struggle with remote-sensing videos, but a new framework boosts their accuracy by over 9% in critical tasks.
Forget training separate models for every remote sensing modality pair: Any2Any learns a single latent space for unified translation, even generalizing to unseen modality combinations.
