About this document
Progressive Active Learning for SIRST Detection by fanwu0409 is a document available to read on EtoBox.
The document presents a Progressive Active Learning (PAL) framework designed to enhance infrared small target detection with single point supervision, addressing issues of instability and excessive label evolution in existing methods. The PAL framework promotes a learning process from easy to hard samples, incorporating a model pre-start concept and a dual-update strategy for effective sample selection and label refinement. Extensive experiments demonstrate that the PAL framework achieves state-of-the-art r
- Author
- fanwu0409
- Language
- EN