change semantic vector embedding to be based on tokenized image only instead of clustering blobs that are the exact same, just represent with 1 blob with some scaling diameter clean up this dashboard, prettify, and add connective lines