As we have seen, FG-Net dataset is so small that cannot be used for training and testing with our current methods. So we are going to try to use the following methodology:
Use CACD dataset to both train and test
Use FG-Net to only test the results of the model
The second point can be really challenging as, shown in EDA, the two datasets have very different age and number of images per individual distributions
As we have seen, FG-Net dataset is so small that cannot be used for training and testing with our current methods. So we are going to try to use the following methodology:
The second point can be really challenging as, shown in EDA, the two datasets have very different age and number of images per individual distributions