Search papers, labs, and topics across Lattice.
Harbin Institute of Technology (Shenzhen)
4
0
5
0
Key contribution not extracted.
ViT-based stereo matching can rival CNNs in detail prediction and generalization to arbitrary resolutions with the right architectural choices.
Concept erasure in text-to-image models is mostly smoke and mirrors: a text-free attack can still regenerate "forgotten" concepts by exploiting the model's latent visual knowledge.
By dynamically adapting activation functions at the pixel level, PAKAN achieves state-of-the-art performance in pansharpening, demonstrating the power of adaptive neural operators for image fusion.
