Disclosed is an unsupervised video object segmentation apparatus and method using attention-based object segmentation. In this embodiment, mask candidates and saliency are generated from a video frame, and similarity between a mask candidate and a registered object is determined based on a Re-ID (Re-Identification) characteristic inferred from the mask candidate. To perform object division. In addition, it provides an unsupervised video object segmentation (UVOS: Unsupervised Video Object Segmentation) apparatus and method capable of discriminating a mask candidate other than a registered object as a new registered object based on the degree of attention, but maintaining the registered object within a certain number.
展开▼