Stanford Dogs dataset for Fine-Grained Visual Categorization