An Unsupervised Object Localization Based on Topological Data Analysis
Hong Cheng · 2022
In this paper, an unsupervised object localization based on topological image processing is proposed to deal with image localization problem. We use the method to process images with the intention to identify the objects of interest and draw bounding boxes to surrounding the objects in the images. We experiment with a dataset with 200 bird species of 11788 annotated images. Our results show in average 72.41% of our predicted bounding boxes are overlapped with annotated bounding boxes.