Spatial-aware and Supervised Contrastive Representation Learning for Few-Shot Object Detection | AMiner