Abstract
Weakly-supervised semantic segmentation is a challenging task as it outputs pixel-level predictions from weaker labels. Segmentation with weaker labels is an important research area since it can significantly reduce human annotation efforts by associating high-level semantic to low-level appearance. In this article, we propose a novel Regional location Cutting and Dynamic credible regions Correction (RCDC) approach for weakly-supervised semantic segmentation. Only image-level labels are needed and it can take less time for manual annotation. Starting with the weak localization of classification network, our cutting approach combines the weak coverage with the traditional cutting method to obtain the pseudo-labels of around 50% ground truth. Then, our dynamic credible regions correction approach adjusts the loss function during the training to preserve the regions that have the superior performance of each iteration. It can further enhance the pseudo-labels for better segmentation results. Finally, with the fully-connected CRF and the retraining method, our approach obtains a competitive performance on the PASCAL VOC 2012 dataset.
Highlights
With the development of Convolutional Neural Networks (CNNs) recently, computer vision research has made immense progress to improve our lives
By following the two principles we proposed above, we propose a weakly supervised semantic image segmentation method with Regional location Cutting and Dynamic credible regions Correction (RCDC)
Because image-level weakly-supervised semantic segmentation only has the classification labels, most of the segmentation methods are based on Classification Activation Maps (CAMs) which can identify a region of the ground truth using the heatmap of the classification network
Summary
With the development of Convolutional Neural Networks (CNNs) recently, computer vision research has made immense progress to improve our lives. If we can realize semantic segmentation just with weaker labels such as image-level and bounding box labels, a lot of manpower and timing costs will be saved This is of great significance to the engineering application of semantic segmentation. A great number of researchers are focusing on training the segmentation network with unlabelled or weakly labeled images [15]–[18]. With our weakly-supervised image segmentation, we can reach the segmentation results just using image-level labels or bounding box labels. It can save significant time for manual annotation. The dynamic credible regions training correction method based on the dynamic region loss function is developed to correct the wrong regions of the pseudolabels It can improve the results of weakly-supervised segmentation. Fi(c) is got from the unnormalized scores Fi(c) that output from the network through a softmax unit
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
