V2 dataset images about dataset

[rkrasin]

@mayankbhagya same images; more image-level labels; bounding boxes are added.
One important (and a bit unfortunate) change: v2 validation + test set = v1 validation set

@nalldrin please, correct me, if I am wrong.

[mayankbhagya]

I think you're right @rkrasin
Just came across the official announcement post.
From the post:

Today, we introduce an update to Open Images, which contains the addition of a total of ~2M bounding-boxes to the existing dataset

[nalldrin]

Indeed we split the validation set into validation and test. The reason is that 167k images is too large to easily compute metrics during training.

Re: images, the set of images is identical to the v1 release.