Learning from Weak and Noisy Labels for Semantic Segmentation,A weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these `free’ tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. 

 

In this paper, we have proposed a novel approach to learning a semantic segmentation model from both weak and noisy labels. The weakly supervised semantic segmentation problem is cast into a noise reduction problem and a superpixel label noise reduction model is developed based on a novel sparse learning model with an efficient optimisation algorithm. Extensive experiments are carried out to demonstrate that the proposed method is superior to the state-of-the-art methods and alternative sparse learning based label denoising models, particularly when the weak labels are also noisy. 

 

 It is possible to integrate the two into a single model , even though solving it often involves an alternating processing similar to ours.An incremental learning variant of the current method is thus part of the ongoing work.Current efforts thus also include the generalisation of the proposed model to solve a wider range of computer vision problems