Dataset processing issues #48
Description
Hey, I noticed that your processed dataset includes nearly 3,000 more images in the training set and around 6,000 images in the validation set compared to the original RAF-DB. Could you explain how the extra images were generated in your code? I looked through the data_augmentation project you uploaded, and it appears that augmentation is only applied to emotion categories with fewer than 1,000 images, which doesn't seem sufficient to generate such a large number of additional images. I'm sincerely curious about how you achieved this. Were these extra images generated by incorporating images from other datasets, or did you use other techniques to generate them? Additionally, the size of your validation set seems unusual—how did you end up with so many validation images? Please provide a detailed explanation. Thank you!