Introducing a new CoSOD dataset, Co-Saliency of ImageNet (CoSINe), and a novel approach, Hierarchical Instance-aware COnsensus MinEr (HICOME), significantly improves performance in detecting co-salient objects.
Enhancing Co-Salient Object Detection Against Adversarial Attacks
This research paper introduces CONDA, a novel deep learning model for Co-Salient Object Detection (CoSOD) that leverages deep association learning and correspondence-induced association condensation to effectively capture inter-image relationships and improve the accuracy and efficiency of co-salient object detection.