issue of the workflow for similarity graph build and featrue extraction

[json87]

Hi, thanks for providing the toolkit. I find that the workflow for similarity graph build and feature extraction is not correct as features are required for vocabulary tree-based image retrieval. When the database is empty, the step 1 does nothing.

[AIBluefisher]

Yes, you're right. If we use the internal similarity computation method, we must perform feature extraction first.