Ajmal Mian

Brain tumor segmentation can now maintain accuracy even when crucial MRI data is missing, thanks to a novel diffusion-based imputation strategy.

Injecting semantic knowledge from vision transformer embeddings into guided depth super-resolution models dramatically reduces artifacts and sharpens depth boundaries.

Decoupling spectral reflectance and vegetation indices in a two-stream network beats standard multispectral segmentation, especially when generalizing to new fields and years of barley crops.