This visualization effectively demonstrates the efficacy of our tokenizer in accurately capturing and representing the semantic content of the images within the considered dataset and shows its ability to recognize and group image patches that share common visual features and semantic meaning. As it can...