diffrent between single-image-prediction and model-code file

[yuyonod]

if I understand your work, correctly, thrre are no output precision difference between single-image-prediction and model-code is that right? If so I tried to our data-set to predict using afad-coral.py also card-coral.py in single-image-prediction, however output result shows average 10-20 lower age than actual age. Do you have any suggestion to improve precision? Thank you in advance.

[rasbt]

Hi there,

prediction, however output result shows average 10-20 lower age than actual age.

Hm, this sounds weird. It may be there is some issue with that, this shouldn't happen except for some outlier cases maybe. Are you maybe calling the "afad-coral.py " model on CACD images and the "cacd-coral.py" model on AFAD images?

[melih-unsal]

I actually test on my own image instead of the images in the specific dataset. Is it normal to get this kind of gap between the result and ground truth only because I use an image not in the given datasets?