Rethinking Attention: Exploring Shallow Feed-Forward Neural Networks as an Alternative to Attention Layers in Transformers
Paper • 2311.10642 • Published • 25
None defined yet.
Unified Panoramic Geometry Estimation via Multi-View Foundation Models
StereoSpace: Depth-Free Synthesis of Stereo Geometry via End-to-End Diffusion in a Canonical Space