Applying to face recognition

[soroushhashemifar]

Hi. Thank you for this amazing idea on RAG + ImCap. I am wondering if the same approach is applicable to face recognition. Is it possible to have an external database of face representations and let the captioning model use their corresponding names if they exist in the image, or ignore them and use "a person ...", "a man ...", "a woman ...", as alternative? Any idea on that? Thanks in advance for your help.