Closed waxnkw closed 3 years ago
I know that some papers use 5,000 images in the training set as the validation set. In 20 class settings, there are only around 10,000 images which contain 20 relationships used for classification. How many images do you use for validation?
I found the download link of split.npy in scripts/get_visualgenome.sh.
split.npy
scripts/get_visualgenome.sh
I know that some papers use 5,000 images in the training set as the validation set. In 20 class settings, there are only around 10,000 images which contain 20 relationships used for classification. How many images do you use for validation?