If you have ever hit “slice,” watched a forest of supports appear, and shrugged like it is just the price of doing business, you are not alone. Auto supports are convenient, and that convenience can ...
Abstract: Accurate 3D medical image segmentation is crucial for diagnosis and treatment. Diffusion models demonstrate promising performance in medical image segmentation tasks due to the progressive ...
Abstract: 3D instance segmentation (3DIS) aims to identify object instances in a 3D scene by predicting binary foreground masks with corresponding semantic labels. Transformer-based methods have ...
Open-world promptable 3D semantic segmentation remains brittle as semantics are inferred in the input sensor coordinates. Yet humans interpret parts via functional roles in a canonical space – wings ...
Overview of CAFSA-Net. A 3D volume ($C,D,H,W$) is encoded by stacked AsCoT-Former, which apply parallel H/W axial self-attention, sequential D-axis attention, and 3D ...