Builtin Foreground/Background Prior for WeaklySu

Authors: Fatemehsadat Saleh Mohammad Sadegh Aliakbarian Mathieu Salzmann Lars Petersson Stephen Gould Jose M Alvarez

Publish Date: 2016/10/8

Volume: , Issue: , Pages: 413-432

PDF Link

Abstract

Pixellevel annotations are expensive and time consuming to obtain Hence weak supervision using only image tags could have a significant impact in semantic segmentation Recently CNNbased methods have proposed to finetune pretrained networks using image tags Without additional information this leads to poor localization accuracy This problem however was alleviated by making use of objectness priors to generate foreground/background masks Unfortunately these priors either require training pixellevel annotations/bounding boxes or still yield inaccurate object boundaries Here we propose a novel method to extract markedly more accurate masks from the pretrained network itself forgoing external objectness modules This is accomplished using the activations of the higherlevel convolutional layers smoothed by a dense CRF We demonstrate that our method based on these masks and a weaklysupervised loss outperforms the stateoftheart tagbased weaklysupervised semantic segmentation techniques Furthermore we introduce a new form of inexpensive weak supervision yielding an additional accuracy boost

Keywords: