A Comparative Study on Active Learning for Object Detection

Active learning aims to improve the performance of the task model by selecting the mostinformative samples with a limited budget. In object detection tasks, due to high cost of drawingobject bounding boxes (i.e. labeling) for millions of images, effective active learning is crucial.This paper aims to provide a description of two important recent proposed active learningmethods for object detection. Then uses ALBench, a new benchmark for evaluation of activelearning methods and compares these two methods.

کلیدواژه ها:

Computer Vision ، Deep Learning ، Object Detection ، Active Learning