Search papers, labs, and topics across Lattice.
3
0
6
By explicitly disentangling degradation and semantic features with wavelet attention, CWP-Net achieves superior all-in-one image restoration, outperforming previous methods hampered by spurious correlations and biased degradation estimation.
By using optimal transport to guide cross-attention, SceneTransporter disentangles image patches and 3D latents, leading to more coherent and geometrically faithful 3D scene generation from single images.
Ditch density estimation in maximum entropy RL: FLAC leverages kinetic energy regularization within a Schrödinger Bridge framework to achieve state-of-the-art performance without ever needing action log-densities.
