SANSA unlocks the hidden semantics of Segment Anything 2, turning it into a powerful few-shot segmenter for both objects and parts. 🚀 No fine-tuning of SAM2 weights. 🧠️ Fully promptable: points · ...

Prior works refine intermediate attention but face limitations: (1) improvements may not propagate to final segmentation; (2) attention lacks direct class information. We introduce a feedback-driven ...

Abstract: In recent years, deep learning methods such as convolutional neural network (CNN) and transformers have made significant progress in CT multi-organ segmentation. However, CT multi-organ ...

Abstract: Referring Image Segmentation (RIS), aims to segment the object referred by a given sentence in an image by understanding both visual and linguistic information. However, existing RIS methods ...