Due to the lack of shuffling, only the latest (like, last 20% of the images) data becomes the validation data, which is kinda bad.
Due to the lack of shuffling, only the latest (like, last 20% of the images) data becomes the validation data, which is kinda bad.