ELDET: Early-Learning Distillation with Noisy Labels for Object Detection

The performance of learning-based object detection algorithms, which attempt to both classify and locate objects within images, is determined largely by the quality are prev of alent: the annotated objects that dataset are incorrectly used for training.