microCLIP is a lightweight self-training framework that adapts CLIP for fine-grained image classification without requiring labeled data. While CLIP is strong in zero-shot transfer, it primarily ...

Abstract: Unsupervised domain adaptive semantic segmentation (UDA-SS) for remote sensing imagery remains challenging due to the substantial distribution shifts across regions and sensors. Existing ...

Abstract: Underground environments present challenges for 3D point cloud segmentation, including complex structural characteristics and limited semantic annotations. To address these issues, this ...