Training Object Class Detectors with Click Supervision
about 15 days ago.
Updated about 15 days ago.
Having a machine identify an object within an image has made significant progress over the years, however researchers are continuing to work on how to make this happen more accurately and faster, ultimately cheaper. One way of doing that is improving the training data. In their research, we see how the baseline object classification is applied to the image. In most cases, it covers a part of the intended target — but often missing key details. By adding a modest annotation input from humans, in this case asking someone to click on the object, they’ve been able to reduce total annotation time by approximately 10x.
Full Paper: https://arxiv.org/pdf/1704.06189.pdf
Abstract: "Training object class detectors typically requires a large set of images with objects annotated by bounding boxes. However, manually drawing bounding boxes is very time consuming. In this paper we greatly reduce annotation time by proposing center-click annotations: we ask annotators to click on the center of an imaginary bounding box which tightly encloses the object instance. We then incorporate these clicks into existing Multiple Instance Learning techniques for weakly supervised object localization, to jointly localize object bounding boxes over all training images. Extensive experiments on PASCAL VOC 2007 and MS COCO show that: (1) our scheme delivers high-quality detectors, performing substantially better than those produced by weakly supervised techniques, with a modest extra annotation effort; (2) these detectors in fact perform in a range close to those trained from manually drawn bounding boxes; (3) as the center-click task is very fast, our scheme reduces total annotation time by 9× to 18×."