search

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.
Apache License 2.0
4.5k stars 296 forks source link

how to train FaceID Portrait #283

Open jeeveenn opened 5 months ago

jeeveenn commented 5 months ago

hi, what is the difference between faceid-Portrait and faceid in training? does portrait mode require multiple facial images of the same person during training?

xiaohu2015 commented 5 months ago

https://github.com/tencent-ailab/IP-Adapter/issues/266

jeeveenn commented 5 months ago

Thank you for your reply. According to the description in #266, the training methods of faceid-portrait and faceid are the same, except that the training set of faceid-portrait is all images containing faces. Is that so?

xiaohu2015 commented 5 months ago

Thank you for your reply. According to the description in #266, the training methods of faceid-portrait and faceid are the same, except that the training set of faceid-portrait is all images containing faces. Is that so?

we used cropped face images (with 2x face detection bounding boxs)

jeeveenn commented 5 months ago

Thank you for your reply. I'm sorry that I expressed it incorrectly. Faceid-portrait supports the use of multiple facial images for inference . Does it also use multiple images in one inference during the training phase, similar to the way of photomaker?

xiaohu2015 commented 5 months ago

one 1 image in training stage

jeeveenn commented 5 months ago

one 1 image in training stage

Thank you very much for your quick reply!