# Efficient Visual Modeling

MambaVision is a hybrid Mamba-Transformer visual backbone network designed for vision applications, combining the strengths of the Mamba formula and vision Transformers, delivering outstanding performance in image classification and downstream vision tasks.

MambaVision is the first hybrid computer vision model combining the advantages of Mamba and Transformer, significantly enhancing the modeling capability of long-range spatial dependencies through redesigned Mamba formulas and integrated ViT modules.

Hiera is an efficient hierarchical Transformer architecture that optimizes spatial bias characteristics through MAE training, significantly improving parameter utilization efficiency