Search papers, labs, and topics across Lattice.
6
0
6
2
Current models falter on fine-grained facade elements, achieving only 33 IoU across geographic domains, highlighting a pressing need for better benchmarks.
Triplet recall improves by 77.4% with DeWorldSG, setting a new benchmark for 3D scene graph generation.
PANY outperforms existing model-free methods by over 20% in pose accuracy, even in challenging conditions with limited reference overlap.
Thermal infrared alone can drive surprisingly effective 3D Gaussian Splatting, rivaling multi-modal approaches while slashing training time in half.
Ditch ambiguous 2D foundation models: GS4City uses city-model priors to boost semantic Gaussian Splatting, achieving up to 15.8 IoU improvement in coarse building segmentation.
Unlock long-range scene dynamics prediction with Node-RF, a Neural ODE-based NeRF that extrapolates beyond observed trajectories at constant memory cost.
