Salient Object
Salient object detection (SOD) focuses on identifying the most visually prominent objects within an image or video, aiming to mimic human visual attention. Current research emphasizes improving accuracy and efficiency, exploring diverse modalities (RGB, depth, thermal) and incorporating transformer-based architectures, large language models, and diffusion models to enhance feature extraction and reasoning capabilities. This field is significant for applications ranging from image and video analysis to robotics and assistive technologies for the visually impaired, with ongoing efforts to improve robustness and address challenges like camouflaged objects and diverse data types (point clouds, remote sensing imagery).
Papers
Attention2Minority: A salient instance inference-based multiple instance learning for classifying small lesions in whole slide images
Ziyu Su, Mostafa Rezapour, Usama Sajjad, Metin Nafi Gurcan, Muhammad Khalid Khan Niazi
Sharp Eyes: A Salient Object Detector Working The Same Way as Human Visual Characteristics
Ge Zhu, Jinbao Li, Yahong Guo