This dataset contain approx 07 images with well captioned txt file . It can be used to train for flux