International Conference on Computer and Knowledge Engineering

Home / 12th International Conference on Computer and Knowledge Engineering

Semantic Segmentation Using Region Proposals and Weakly-Supervised Learning

Authors :

Maryam Taghizadeh¹ Abdolah Chalechale²

1- Razi University 2- Razi University

Keywords :

Semantic segmentation،Weakly-supervised learning،Region proposal

Abstract :

Region proposal plays an important role in computer vision and successfully improves performance. This paper presents an efficient method using the region proposal for semantic segmentation. The main aim is to generate annotated data for deep learning techniques effortlessly. For this purpose, a region proposal algorithm is used to convert an image into several regions. According to defined rules, regions are explored, and some precise regions are selected. A new algorithm is introduced to generate useful masks only by supervising annotated data in the form of the bounding box. After that, these masks are fed to a deep semantic segmentation network. The proposed method shows good results for weakly supervised learning semantic segmentation on the VOC2012 dataset. Also, this method can be employed to generate huge annotated data automatically and used to train deep networks.