Search papers, labs, and topics across Lattice.
4
0
3
MambaPSA achieves a remarkable 17.6% boost in CPU inference speed while maintaining competitive accuracy in YOLO26, showcasing the potential of state space models in object detection.
Reload-Mamba achieves a remarkable 2.2 mIoU improvement over previous models by specifically addressing response dilution in multi-class semantic segmentation.
ERN-Net achieves superior document binarization performance in low-data and low-memory scenarios, making it a game-changer for resource-constrained applications.
ResNet-based segmentation can still achieve state-of-the-art results without transformers, proving that clever receptive field fusion can rival complex attention mechanisms.
