for natural language processing and data mining
Challenge 1: Train on any dataset, test your method with 1 million distractors Challenge 2: Training on 672K identities (4.7 Million photos), test at Million scale