problems about preprocessing on emotionet

[DHPO]

1. about image: how to crop and resize image? I tried to use openface's alignment, the training result is good, but when apply on wild image, the model cannot tolerant any unalignment of image, slight rotation will cause bad result; so I want to use face_recognition to preprocess (same as test.py), but haven't tried yet. And is there any way else (or something I should pay attention to) to do so?
2. about AU: do you use AU of dataset or the output of openface? I tried the latter and got a good result. And will the former give a better result?

By the way, if I want to change a single AU, should I just leave others 0 or fill others by the openface AU output?

Thanks in advance

[octaviawfx]

1. about image: how to crop and resize image? I tried to use openface's alignment, the training result is good, but when apply on wild image, the model cannot tolerant any unalignment of image, slight rotation will cause bad result; so I want to use face_recognition to preprocess (same as test.py), but haven't tried yet. And is there any way else (or something I should pay attention to) to do so?
2. about AU: do you use AU of dataset or the output of openface? I tried the latter and got a good result. And will the former give a better result?

By the way, if I want to change a single AU, should I just leave others 0 or fill others by the openface AU output?

Thanks in advance

HI，when you tried to get AU, you give OPENFACE original pictures, but give the model cropped images as input, am I right?