Work

High-Order Multi-Scale Attention and Vertical Discriminator Enhanced CLIP for Monocular Depth Estimation